# List of open-weight models

> Source: https://aiwiki.ai/wiki/list_of_open_weight_models
> Updated: 2026-11-17
> Fact-checked: 2026-11-17
> Categories: AI Models, Open Source AI
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) - attribute to "AI Wiki (aiwiki.ai)"
> Cite as: AI Wiki. "List of open-weight models." aiwiki.ai, 17 Nov 2026. https://aiwiki.ai/wiki/list_of_open_weight_models
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution.

This is a list of **open-weight models**: artificial intelligence models whose trained parameters ([weights](https://aiwiki.ai/wiki/open_weights)) the developer has published for download, so that the model can be run, studied and in most cases fine-tuned outside the developer's own service. It covers the 408 models and model families that have an AI Wiki article recording downloadable weights as of November 17, 2026: 374 whose weights are available in full and 34 families in which only some members are, marked "part of the family" below. Models that are offered only through a hosted service are not listed, and the list is not exhaustive: it is limited to models with their own article. The same data can be searched and filtered by type, license and memory in the [open-weight model directory](https://aiwiki.ai/tools/open_models).

"Open weights" describes what can be downloaded, not the terms of use. The licenses in this list range from standard permissive software licenses to research-only terms, so each row gives the license of the weights and whether it allows commercial use. A license on the weights does not by itself make a model open source in the sense of the [Open Source Initiative](https://aiwiki.ai/wiki/open_source_initiative)'s [Open Source AI Definition](https://aiwiki.ai/wiki/open_source_ai_definition), which asks for the source code used to train and run the system and detailed information about the training data in addition to the parameters.[1]

## How to read the list

Each row summarizes the fact box of the linked article, where the license file, model card or announcement behind it is cited. The columns mean the following.

- **Released** is the release date given in the linked article; for a family it is usually the date of the first member. For some 2026 models it is the date of the hosted launch, and the weights followed days or weeks later; the linked article gives both dates.
- **Parameters** is the size stated by the developer, as recorded in the linked article. Where a model activates only part of its parameters for each token, both the total and the active count are given when the developer states them.
- **Weights license** is the license of the downloadable weights, which can differ from the license of the accompanying code. For a family it is normally the license of the most recent open release, and for a release with several sizes the license that covers most of them; the linked article notes the exceptions.
- **Commercial use** is "Yes" when the license allows commercial use with no conditions beyond notices, attribution or share-alike terms; "With conditions" when commercial use is allowed subject to conditions such as revenue or user thresholds, an acceptable use policy, territorial exclusions or naming and display requirements; "No" when the license covers research or non-commercial use only; and "Not stated" when the repository carries no license statement for the weights.

Within each section the newest releases come first.

## Licenses at a glance

The table counts the 408 entries by their Commercial use value. A partly open family is counted under the value of its open members.

| Commercial use | Entries | Typical licenses |
|---|---|---|
| Yes | 238 | Standard licenses such as [Apache 2.0](https://aiwiki.ai/wiki/apache_license), [MIT](https://aiwiki.ai/wiki/mit_license) and CC BY 4.0, which allow commercial use, modification and redistribution subject to notice or attribution terms [2][3][4]; also the BSD licenses and [OpenMDW](https://aiwiki.ai/wiki/openmdw) |
| With conditions | 113 | Licenses written by the developer, such as the Llama Community License, the NVIDIA Open Model License and model licenses with revenue or user thresholds, and the [OpenRAIL](https://aiwiki.ai/wiki/openrail) licenses, which restrict certain uses; see [open-weight license comparison](https://aiwiki.ai/wiki/open_weight_license_comparison) |
| No | 55 | Terms such as CC BY-NC 4.0, which does not license commercial use [5], and developers' research licenses |
| Not stated | 2 | Repositories with no license statement for the weights |

Of the 408 entries, 34 are partly open families: some of their models have downloadable weights, and the others are hosted only or were never released. The most common licenses are Apache 2.0 (139), MIT (71), CC BY-NC 4.0 (26), Llama Community License (15), OpenMDW 1.1 (11), NVIDIA Open Model License (11), Gemma Terms of Use (10), CC BY 4.0 (10).

## Language models

### 2026

| Model | Developer | Released | Parameters | Weights license | Commercial use |
|---|---|---|---|---|---|
| [ERNIE 5.5 Flash Base](https://aiwiki.ai/wiki/ernie_5_5_flash_base) | Baidu | November 5, 2026 | 280B total, 14B active | ERNIE 5.5 Flash Base Model License Agreement v1.0 | With conditions |
| [LongCat-2.5](https://aiwiki.ai/wiki/longcat_2_5) | Meituan LongCat team | November 2, 2026 | 1.6T total, about 48B active | MIT | Yes |
| [Ripple-65B Preview](https://aiwiki.ai/wiki/ripple_65b_preview) | Ricursive Intelligence | November 2026 | About 65B | Apache 2.0 | Yes |
| [LFM3](https://aiwiki.ai/wiki/lfm3) | Liquid AI | October 27, 2026 | 1.73B to 2.17B | LFM Open License v1.0 | With conditions |
| [Kimi K3.1 Pro](https://aiwiki.ai/wiki/kimi_k3_1_pro) | Moonshot AI | October 23, 2026 | 2.8T total, 104B active | Kimi K3.1-Pro License | With conditions |
| [Nemotron 3.5 Bolt E4B](https://aiwiki.ai/wiki/nemotron_3_5_bolt_e4b) | NVIDIA | October 23, 2026 | 8B total (4B effective) | OpenMDW 1.1 | Yes |
| [Recursive Flint](https://aiwiki.ai/wiki/recursive_flint) | Recursive Superintelligence | October 21, 2026 | 39.8B total; approximately 2.5B active per token | Apache 2.0 | Yes |
| [Spark-X2.6-4B](https://aiwiki.ai/wiki/spark_x2_6_4b) | SparkLLM / XHToken | October 21, 2026 | 4,112,079,360 | Apache 2.0 | Yes |
| [Ornith-1.5-122B](https://aiwiki.ai/wiki/ornith_1_5_122b) | Ornith (ornith-ai) | October 20, 2026 | 125.1B in checkpoint metadata (named 122B) | MIT | Yes |
| [MiniMax M3.1 Flash](https://aiwiki.ai/wiki/minimax_m3_1_flash) | MiniMax | October 19, 2026 | About 428B total, 23B active per token | MiniMax Community License | With conditions |
| [Beam (Reflection)](https://aiwiki.ai/wiki/beam_reflection) | Reflection AI | October 16, 2026 | 501.3B total, 23.4B active | Apache 2.0 | Yes |
| [Muse Spark Open](https://aiwiki.ai/wiki/muse_spark_open) | Meta Superintelligence Labs | October 15, 2026 | 816B total, 42B active | Muse Model License | With conditions |
| [Nemotron 3.5 Bolt](https://aiwiki.ai/wiki/nemotron_3_5_bolt) | NVIDIA | October 15, 2026 | 74B total, 9B active | OpenMDW 1.1 | Yes |
| [Nemotron 3.5 Super VL](https://aiwiki.ai/wiki/nemotron_3_5_super_vl) | NVIDIA | October 15, 2026 | 120B total, 12B active per token | OpenMDW 1.1 | Yes |
| [Underdog Saluki 27B](https://aiwiki.ai/wiki/underdog_saluki_27b) | Underdog (Conway Research) | October 7, 2026 | 27B (2-bit quantized; 7.89 GB GGUF file) | CC BY-NC 4.0 | No |
| [EmbeddingGemma 2](https://aiwiki.ai/wiki/embeddinggemma_2) | Google DeepMind | October 6, 2026 | About 740M (all modality encoders loaded) | Apache 2.0 | Yes |
| [Mistral Large 4](https://aiwiki.ai/wiki/mistral_large_4) | Mistral AI | October 6, 2026 | About 1T total, 49B active per token | Mistral Ecosystem Open-Weight License | With conditions |
| [Kolibri 1](https://aiwiki.ai/wiki/kolibri_1) | Aleph Alpha | October 3, 2026 | 78.1B total, 3.46B active (excluding embeddings) | Apache 2.0 | Yes |
| [Alice AI Foundation 80B-A3B](https://aiwiki.ai/wiki/alice_ai_foundation_80b_a3b) | Yandex | September 21, 2026 | 80B total, 3B active | Apache 2.0 | Yes |
| [Xiaomi MiMo-V2.6](https://aiwiki.ai/wiki/mimo_v2_6) | Xiaomi (MiMo team) | September 21, 2026 | 310B total, 15B active (Flash) | MIT | Yes |
| [Step 5 Preview](https://aiwiki.ai/wiki/step_5_preview) | StepFun | September 20, 2026 | 600B total, 27B active | StepFun Model License 1.0 | With conditions |
| [DeepSeek V4.1-Flash](https://aiwiki.ai/wiki/deepseek_v4_1_flash) | DeepSeek | September 10, 2026 | Not stated | MIT | Yes |
| [MiniCPM5-2B](https://aiwiki.ai/wiki/minicpm5_2b) | OpenBMB | September 7, 2026 | 2.5B (2,516,756,480 total; 1,981,982,720 non-embedding) | Apache 2.0 | Yes |
| [Hy4 Preview](https://aiwiki.ai/wiki/hy4_preview) | Tencent Hy Team | August 28, 2026 | 770B total, 49B active (backbone) | Apache 2.0 | Yes |
| [GLM-5.3-Flash](https://aiwiki.ai/wiki/glm_5_3_flash) | Z.ai | August 26, 2026 | 320B total, 18B active | MIT | Yes |
| [Qwen3.8-Flash-Next](https://aiwiki.ai/wiki/qwen3_8_flash_next) | Qwen team, Alibaba | August 26, 2026 | 125B main model, 6B active | Qwen Community License 1.0 | With conditions |
| [GLM-5.3](https://aiwiki.ai/wiki/glm_5_3) | Z.ai | August 14, 2026 | 744B total, 40B active (Z.ai model-index label) | GLM-5.3 License | With conditions |
| [dots3-note Preview](https://aiwiki.ai/wiki/dots3_note) | dots studio (Xiaohongshu) | August 13, 2026 | 280B total, 16B active (language backbone) | Apache 2.0 | Yes |
| [Nemotron 3.5 Lightning](https://aiwiki.ai/wiki/nemotron_3_5_lightning) | NVIDIA | August 11, 2026 | 30B total, 3B active (31.6B / 3.6B counting embeddings) | OpenMDW 1.1 | Yes |
| [Muse Glimmer](https://aiwiki.ai/wiki/muse_glimmer) | Meta AI | August 10, 2026 | About 29.6B (29,776,626,688 in the BF16 checkpoint) | Apache 2.0 | Yes |
| [Qwen3.8-Max](https://aiwiki.ai/wiki/qwen3_8_max) | Qwen team, Alibaba | August 3, 2026 | 2.4T total, 95B active | Qwen3.8-Max License | With conditions |
| [BigBang-v1](https://aiwiki.ai/wiki/bigbang_v1) | The Endless Frontier team | August 2, 2026 | 35B total, about 3B active | Apache 2.0 | Yes |
| [Ling-3.0-flash](https://aiwiki.ai/wiki/ling_3_0_flash) | inclusionAI | July 23, 2026 | 124B total, 5.1B active per token | MIT | Yes |
| [Qwen3.8](https://aiwiki.ai/wiki/qwen3_8) | Qwen team, Alibaba Cloud | July 19, 2026 | 27.8B (Qwen3.8-27B); 2.4T total, 95B active (Qwen3.8-Max) | Qwen3.8-Max License | With conditions |
| [Kimi K3](https://aiwiki.ai/wiki/kimi_k3) | Moonshot AI | July 16, 2026 | 2.78T total; 104.2B activated per token | Kimi K3 License | With conditions |
| [Inkling (Thinking Machines Lab)](https://aiwiki.ai/wiki/inkling) | Thinking Machines Lab | July 15, 2026 | 975B total, 41B active | Apache 2.0 | Yes |
| [Leanstral 1.5](https://aiwiki.ai/wiki/leanstral_1_5) | Mistral AI | June 30, 2026 | 119B total, 6.5B active | Apache 2.0 | Yes |
| [Kimi K2.7-Code](https://aiwiki.ai/wiki/kimi_k2_7_code) | Moonshot AI | June 12, 2026 | 1T total, 32B active | Modified MIT | With conditions |
| [DiffusionGemma](https://aiwiki.ai/wiki/diffusiongemma) | Google DeepMind | June 10, 2026 | 25.2B total, 3.8B active | Apache 2.0 | Yes |
| [North Mini Code](https://aiwiki.ai/wiki/north_mini_code) | Cohere | June 9, 2026 | 30B total, about 3B active | Apache 2.0 | Yes |
| [MiniMax M3](https://aiwiki.ai/wiki/minimax_m3) | MiniMax | June 1, 2026 | Not stated | MiniMax Community License | With conditions |
| [GLM-5.2](https://aiwiki.ai/wiki/glm_5_2) | Zhipu AI (Z.ai) | June 2026 | 744B total, 40B active per Z.ai's model index | MIT | Yes |
| [Hy Memory](https://aiwiki.ai/wiki/hy_memory) | Tencent Hunyuan (Tencent Hy) | May 28, 2026 | About 3B active (Hy-Memory-30B-A3B) | Apache 2.0 | Yes |
| [ZAYA1-8B](https://aiwiki.ai/wiki/zaya1_8b) | Zyphra | May 6, 2026 | 8.4B total; 760M active per token | Apache 2.0 | Yes |
| [Mistral Medium 3.5](https://aiwiki.ai/wiki/mistral_medium_3_5) | Mistral AI | April 28, 2026 | ~128 billion (all active per token) | Modified MIT | With conditions |
| [DeepSeek V4](https://aiwiki.ai/wiki/deepseek_v4) | DeepSeek | April 24, 2026 | 1.6T total, 49B active (Pro); 284B total, 13B active (Flash) | MIT | Yes |
| [DeepSeek V4-Flash](https://aiwiki.ai/wiki/deepseek_v4_flash) | DeepSeek | April 24, 2026 | 284B total, 13B active | MIT | Yes |
| [DeepSeek V4-Pro](https://aiwiki.ai/wiki/deepseek_v4_pro) | DeepSeek | April 24, 2026 | 1.6T total, 49B active | MIT | Yes |
| [Tencent Hunyuan Hy3](https://aiwiki.ai/wiki/hunyuan_hy3) | Tencent | April 23, 2026 | ~295B total; ~21B active per token | Apache 2.0 | Yes |
| [Kimi K2.6](https://aiwiki.ai/wiki/kimi_k2_6) | Moonshot AI | April 20, 2026 | ~1T total; ~32B active per token | Modified MIT | With conditions |
| [Muse Spark](https://aiwiki.ai/wiki/muse_spark) (part of the family) | Meta Superintelligence Labs | April 8, 2026 | 816B total, 42B active (Muse Spark Open) | Muse Model License | With conditions |
| [GLM-5.1](https://aiwiki.ai/wiki/glm_5_1) | Zhipu AI (Z.ai) | April 7, 2026 | 744B total, 40B active per Z.ai's model index | MIT | Yes |
| [Gemma 4](https://aiwiki.ai/wiki/gemma_4) | Google DeepMind | April 2, 2026 | E2B 5.1B; E4B 8B; 26B A4B 25.2B (3.8B active); 31B 30.7B | Apache 2.0 | Yes |
| [Qwen3.6](https://aiwiki.ai/wiki/qwen3_6) (part of the family) | Alibaba (Qwen team) | April 2026 | Open: 35B/3B active (MoE), 27B (dense); hosted undisclosed | Apache 2.0 | Yes |
| [MiniMax M2.7](https://aiwiki.ai/wiki/minimax_m2_7) | MiniMax | March 18, 2026 | ~230B total; ~10B active per token | MiniMax M2.7 Non-Commercial License | No |
| [Mistral Small 4](https://aiwiki.ai/wiki/mistral_small_4) | Mistral AI | March 16, 2026 | 119 billion total; ~6.5 billion active per token | Apache 2.0 | Yes |
| [Qwen3.5](https://aiwiki.ai/wiki/qwen3_5) | Qwen team, Alibaba | February 16, 2026 | 397B total | Apache 2.0 | Yes |
| [GLM-5](https://aiwiki.ai/wiki/glm_5) | Zhipu AI (Z.ai) | February 11, 2026 | 744B total; ~40B active per token | MIT | Yes |
| [Qwen3-Coder-Next](https://aiwiki.ai/wiki/qwen3_coder_next) | Alibaba (Qwen team) | February 3, 2026 | About 80B total, about 3B active per token | Apache 2.0 | Yes |
| [Kimi K2.5](https://aiwiki.ai/wiki/kimi_k2_5) | Moonshot AI | January 27, 2026 | 1T total; 32B active per token | Modified MIT | With conditions |
| [Jamba2](https://aiwiki.ai/wiki/jamba_2) | AI21 Labs | January 8, 2026 | 3B dense; Mini 52B total, 12B active | Apache 2.0 | Yes |

### 2025

| Model | Developer | Released | Parameters | Weights license | Commercial use |
|---|---|---|---|---|---|
| [Nemotron 3](https://aiwiki.ai/wiki/nemotron_3) | NVIDIA | December 15, 2025 | Nano 30B (3.5B active); Super 120.6B (12.7B) | NVIDIA Open Model License | With conditions |
| [Mistral Large 3](https://aiwiki.ai/wiki/mistral_large_3) | Mistral AI | December 2, 2025 | ~675B total; ~41B active per token (vision encoder ~2.5B) | Apache 2.0 | Yes |
| [OLMo 3](https://aiwiki.ai/wiki/olmo_3) | Allen Institute for AI (Ai2) | November 20, 2025 | 7B and 32B | Apache 2.0 | Yes |
| [Kimi K2 Thinking](https://aiwiki.ai/wiki/kimi_k2_thinking) | Moonshot AI | November 6, 2025 | About 1 trillion total; about 32 billion active per token | Modified MIT | With conditions |
| [Kimi Linear](https://aiwiki.ai/wiki/kimi_linear) | Moonshot AI | October 30, 2025 | 48B total, 3B active | MIT | Yes |
| [MiniMax M2](https://aiwiki.ai/wiki/minimax_m2) | MiniMax | October 27, 2025 | 230B total; ~10B active per token | Modified MIT | With conditions |
| [Ling-1T](https://aiwiki.ai/wiki/ling_1t) | Ant Group (Inclusion AI) | October 9, 2025 | ~1 trillion total; 51B active per token | MIT | Yes |
| [Jamba Reasoning 3B](https://aiwiki.ai/wiki/jamba_reasoning_3b) | AI21 Labs | October 8, 2025 | ~3B | Apache 2.0 | Yes |
| [IBM Granite 4.0](https://aiwiki.ai/wiki/granite_4) | IBM | October 2, 2025 | ~3B (Micro) to ~32B total with ~9B active (H-Small) | Apache 2.0 | Yes |
| [Ring-1T](https://aiwiki.ai/wiki/ring_1t) | inclusionAI (Ant Group) | October 2025 | ~1T total; ~50B active per token | MIT | Yes |
| [GLM-4.6](https://aiwiki.ai/wiki/glm_4_6) | Zhipu AI (Z.ai) | September 30, 2025 | ~357B total; ~32B active per token | MIT | Yes |
| [DeepSeek V3.2](https://aiwiki.ai/wiki/deepseek_v3_2) | DeepSeek | September 29, 2025 | 671B total; 37B active per token | MIT | Yes |
| [Qwen3-VL](https://aiwiki.ai/wiki/qwen3_vl) | Qwen team, Alibaba Cloud | September 23, 2025 | Dense 2B, 4B, 8B, 32B; MoE 30B-A3B and 235B-A22B | Apache 2.0 | Yes |
| [EmbeddingGemma](https://aiwiki.ai/wiki/embeddinggemma) | Google DeepMind | September 4, 2025 | ~308M (~100M backbone, ~200M embedding table) | Gemma Terms of Use | With conditions |
| [Apertus](https://aiwiki.ai/wiki/apertus) | Swiss AI Initiative (EPFL, ETH Zurich, CSCS) | September 2, 2025 | 8B and 70B (70B reported as 71B); base and Instruct variants | Apache 2.0 | Yes |
| [LongCat-Flash](https://aiwiki.ai/wiki/longcat_flash) | Meituan (LongCat team) | September 1, 2025 | ~560B total; 18.6B-31.3B active per token (avg ~27B) | MIT | Yes |
| [Qwen3-Next](https://aiwiki.ai/wiki/qwen3_next) | Qwen team, Alibaba Cloud | September 2025 | 80 billion total; ~3 billion active per token | Apache 2.0 | Yes |
| [Qwen3-Omni](https://aiwiki.ai/wiki/qwen3_omni) | Qwen team, Alibaba Cloud | September 2025 | 30B total (Thinker); ~3B active per token | Apache 2.0 | Yes |
| [Command A Reasoning](https://aiwiki.ai/wiki/command_a_reasoning) | Cohere and Cohere Labs | August 21, 2025 | 111 billion | CC BY-NC 4.0 | No |
| [Jet-Nemotron](https://aiwiki.ai/wiki/jet_nemotron) | NVIDIA Research | August 21, 2025 | ~2B and ~4B variants | NVIDIA License (non-commercial) | No |
| [DeepSeek V3.1](https://aiwiki.ai/wiki/deepseek_v3_1) | DeepSeek | August 19, 2025 | 671B total, 37B active per token | MIT | Yes |
| [Nemotron Nano 2](https://aiwiki.ai/wiki/nemotron_nano_2) | NVIDIA | August 18, 2025 | ~9B (flagship, pruned and distilled from a 12B base) | NVIDIA Open Model License | With conditions |
| [gpt-oss](https://aiwiki.ai/wiki/gpt_oss) | OpenAI | August 5, 2025 | 120b: 117B total, 5.1B active; 20b: 21B total, 3.6B active | Apache 2.0 | Yes |
| [Hermes 4](https://aiwiki.ai/wiki/hermes_4) | Nous Research | August 2025 | 14B, 70B, 405B | Llama Community License | With conditions |
| [FoxBrain](https://aiwiki.ai/wiki/foxbrain) (part of the family) | Hon Hai Research Institute (Foxconn) | July 31, 2025 | 70B (v1.x) | FoxBrain Model Usage and License Agreement | No |
| [Step-3](https://aiwiki.ai/wiki/step_3) | StepFun | July 31, 2025 | ~321B total; ~38B active per token | Apache 2.0 | Yes |
| [GLM-4.5](https://aiwiki.ai/wiki/glm_4_5) | Zhipu AI (Z.ai) | July 28, 2025 | 355B total, 32B active; Air: 106B total, 12B active | MIT | Yes |
| [Qwen3-Coder](https://aiwiki.ai/wiki/qwen_3_coder) | Alibaba Qwen team (Tongyi Lab) | July 22, 2025 | 480B total; 35B active per token (flagship) | Apache 2.0 | Yes |
| [Kimi K2](https://aiwiki.ai/wiki/kimi_k2) | Moonshot AI | July 11, 2025 | 1.04T total; 32.6B active per token | Modified MIT | With conditions |
| [SmolLM 3](https://aiwiki.ai/wiki/smollm_3) | Hugging Face | July 8, 2025 | 3B | Apache 2.0 | Yes |
| [Phi-4-mini-flash-reasoning](https://aiwiki.ai/wiki/phi_4_mini_flash_reasoning) | Microsoft | July 2025 | 3.8 billion | MIT | Yes |
| [MiniMax M1](https://aiwiki.ai/wiki/minimax_m1) | MiniMax | June 16, 2025 | 456B total; 45.9B active per token | Apache 2.0 | Yes |
| [Magistral](https://aiwiki.ai/wiki/magistral) (part of the family) | Mistral AI | June 10, 2025 | 24B (Magistral Small); unpublished (Magistral Medium) | Apache 2.0 | Yes |
| [Hunyuan-A13B](https://aiwiki.ai/wiki/hunyuan_a13b) | Tencent (Hunyuan team) | June 2025 | ~80B total; ~13B active per token | Tencent Hunyuan Community License | With conditions |
| [Qwen3 Embedding](https://aiwiki.ai/wiki/qwen3_embedding) | Alibaba Qwen team | June 2025 | 0.6B, 4B, 8B | Apache 2.0 | Yes |
| [Devstral](https://aiwiki.ai/wiki/devstral) (part of the family) | Mistral AI with All Hands AI | May 21, 2025 | 24 billion (Small); Medium not disclosed | Apache 2.0 | Yes |
| [Falcon-H1](https://aiwiki.ai/wiki/falcon_h1) | Technology Innovation Institute (TII), Abu Dhabi | May 20, 2025 | Six sizes: 0.5B, 1.5B, 1.5B-Deep, 3B, 7B, 34B | TII Falcon License | With conditions |
| [Gemma 3n](https://aiwiki.ai/wiki/gemma_3n) | Google DeepMind | May 20, 2025 | 5B (E2B) and 8B (E4B) raw; about 2B and 4B effective | Gemma Terms of Use | With conditions |
| [Phi-4 Reasoning](https://aiwiki.ai/wiki/phi_4_reasoning) | Microsoft | April 30, 2025 | 14 billion | MIT | Yes |
| [Qwen3](https://aiwiki.ai/wiki/qwen_3) | Qwen Team, Alibaba Cloud | April 29, 2025 | 0.6B to 235B (flagship 235B total; 22B active) | Apache 2.0 | Yes |
| [InternVL3](https://aiwiki.ai/wiki/internvl3) | OpenGVLab (Shanghai AI Laboratory) | April 11, 2025 | Seven dense sizes, 938M to 78.41B | Qwen License (Tongyi Qianwen) | With conditions |
| [Llama 4](https://aiwiki.ai/wiki/llama_4) | Meta | April 5, 2025 | Scout: 109B total, 17B active | Llama Community License | With conditions |
| [Llama 4 Scout and Maverick](https://aiwiki.ai/wiki/llama_4_scout_maverick) | Meta | April 5, 2025 | Scout: ~109B total, 17B active (16 experts) | Llama Community License | With conditions |
| [Apriel (ServiceNow)](https://aiwiki.ai/wiki/apriel) | ServiceNow (AI Research / SLAM lab) | April 2025 | ~4.8B to ~15B | MIT | Yes |
| [Nemotron-H](https://aiwiki.ai/wiki/nemotron_h) | NVIDIA | April 2025 | 8B, 47B, and 56B (base models) | NVIDIA Internal Scientific R&D Model License | No |
| [Llama Nemotron](https://aiwiki.ai/wiki/llama_nemotron) | Nvidia | March 18, 2025 | 8B (Nano); 49B (Super); 253B (Ultra) | NVIDIA Open Model License | With conditions |
| [RWKV-7 (Goose)](https://aiwiki.ai/wiki/rwkv_7) | RWKV Project (Bo Peng et al., with EleutherAI and Recursal AI) | March 18, 2025 | 0.19B to 2.9B (World v3 series); later 7.2B G1 checkpoint | Apache 2.0 | Yes |
| [ERNIE 4.5](https://aiwiki.ai/wiki/ernie_4_5) | Baidu | March 16, 2025 | 0.3B dense up to 424B total (47B active) across ten variants | Apache 2.0 | Yes |
| [Cohere Command A](https://aiwiki.ai/wiki/command_a) | Cohere | March 13, 2025 | 111B | CC BY-NC 4.0 | No |
| [Gemma 3](https://aiwiki.ai/wiki/gemma_3) | Google DeepMind | March 12, 2025 | 1B, 4B, 12B, 27B (270M added August 2025) | Gemma Terms of Use | With conditions |
| [Skywork-R1V](https://aiwiki.ai/wiki/skywork_r1v) | Skywork AI (Kunlun Tech) | March 2025 | 38B (each open generation) | MIT | Yes |
| [Phi-4-mini](https://aiwiki.ai/wiki/phi_4_mini) | Microsoft | February 26, 2025 | 3.8 billion | MIT | Yes |
| [LLaDA (Large Language Diffusion)](https://aiwiki.ai/wiki/llada) | Renmin University of China and Ant Group | February 14, 2025 | 8B (LLaDA 8B); up to 100B (LLaDA 2.0 flash) | MIT | Yes |
| [Goedel-Prover](https://aiwiki.ai/wiki/goedel_prover) | Princeton Language and Intelligence, with Tsinghua, Amazon, Meta FAIR, Numina | February 2025 | 7B (V1); 8B and 32B (V2) | Apache 2.0 | Yes |
| [olmOCR](https://aiwiki.ai/wiki/olmocr) | Allen Institute for AI (Ai2) | February 2025 | 7B | Apache 2.0 | Yes |
| [DeepSeek-R1](https://aiwiki.ai/wiki/deepseek_r1) | DeepSeek | January 20, 2025 | 671B total; 37B active per token | MIT | Yes |
| [DeepSeek-R1-Distill](https://aiwiki.ai/wiki/deepseek_r1_distill) | DeepSeek | January 20, 2025 | 1.5B, 7B, 8B, 14B, 32B, and 70B variants | MIT | Yes |
| [MiniMax-Text-01](https://aiwiki.ai/wiki/minimax_text_01) | MiniMax | January 14, 2025 | 456B total; 45.9B active per token | MiniMax Model License | With conditions |

### 2024

| Model | Developer | Released | Parameters | Weights license | Commercial use |
|---|---|---|---|---|---|
| [DeepSeek V3](https://aiwiki.ai/wiki/deepseek_v3) | DeepSeek | December 26, 2024 | 671B total (about 685B with MTP module) | DeepSeek Model License | With conditions |
| [QvQ](https://aiwiki.ai/wiki/qvq) (part of the family) | Qwen team, Alibaba | December 25, 2024 | 72B (QVQ-72B-Preview) | Qwen License (Tongyi Qianwen) | With conditions |
| [Falcon 3](https://aiwiki.ai/wiki/falcon_3) | Technology Innovation Institute (TII) | December 17, 2024 | 1B, 3B, 7B, 10B (transformer); 7B (Mamba) | TII Falcon License | With conditions |
| [Byte Latent Transformer](https://aiwiki.ai/wiki/byte_latent_transformer) | Meta AI (FAIR) | December 13, 2024 | Up to 8B trained; 1B and 7B checkpoints released | FAIR Noncommercial Research License | No |
| [Phi-4](https://aiwiki.ai/wiki/phi_4) | Microsoft Research | December 12, 2024 | 14 billion (about 14.7B with embeddings) | MIT | Yes |
| [Llama 3.3](https://aiwiki.ai/wiki/llama_3_3) | Meta | December 6, 2024 | 70 billion | Llama Community License | With conditions |
| [QwQ](https://aiwiki.ai/wiki/qwq) | Alibaba Cloud Qwen team | November 28, 2024 | 32.5B (dense) | Apache 2.0 | Yes |
| [OLMo 2](https://aiwiki.ai/wiki/olmo_2) | Allen Institute for AI (Ai2) | November 26, 2024 | 7B, 13B, 32B | Apache 2.0 | Yes |
| [Pixtral Large](https://aiwiki.ai/wiki/pixtral_large) | Mistral AI | November 18, 2024 | ~124 billion total | Mistral Research License | No |
| [SmolLM 2](https://aiwiki.ai/wiki/smollm_2) | Hugging Face | November 1, 2024 | 135M, 360M, 1.7B (three sizes) | Apache 2.0 | Yes |
| [Marco-o1](https://aiwiki.ai/wiki/marco_o1) | Alibaba International Digital Commerce (MarcoPolo team) | November 2024 | Not stated | Apache 2.0 | Yes |
| [Ministral 3B / 8B](https://aiwiki.ai/wiki/ministral) (part of the family) | Mistral AI | October 16, 2024 | ~3B and ~8B (8,019,808,256 for the 8B) | Apache 2.0 | Yes |
| [Llama-3.1-Nemotron-70B-Instruct](https://aiwiki.ai/wiki/llama_3_1_nemotron_70b) | NVIDIA | October 2024 | 70B | NVIDIA Open Model License | With conditions |
| [Llama 3.2](https://aiwiki.ai/wiki/llama_3_2) | Meta | September 25, 2024 | 1.23B, 3.21B (text-only); 11B, 90B (vision) | Llama Community License | With conditions |
| [Llama 3.2 Vision](https://aiwiki.ai/wiki/llama_3_2_vision) | Meta AI | September 25, 2024 | 11B (10.6B actual) and 90B (88.8B actual) | Llama Community License | With conditions |
| [Qwen2.5](https://aiwiki.ai/wiki/qwen2_5) | Alibaba Cloud Qwen team | September 19, 2024 | Seven dense sizes from 0.49B to 72.7B | Apache 2.0 | Yes |
| [Qwen2.5-Math](https://aiwiki.ai/wiki/qwen2_5_math) | Qwen team, Alibaba Cloud | September 19, 2024 | 1.5B, 7B and 72B | Apache 2.0 | Yes |
| [Jina Embeddings v3](https://aiwiki.ai/wiki/jina_embeddings_v3) | Jina AI | September 18, 2024 | 570 million | CC BY-NC 4.0 | No |
| [NVLM](https://aiwiki.ai/wiki/nvlm) | NVIDIA | September 17, 2024 | ~79B (NVLM-D-72B: 72B LLM plus 6B vision encoder) | CC BY-NC 4.0 | No |
| [Pixtral](https://aiwiki.ai/wiki/pixtral) | Mistral AI | September 11, 2024 | ~12.4B (12B: 12B decoder + 400M encoder); 124B (Large) | Apache 2.0 | Yes |
| [OLMoE](https://aiwiki.ai/wiki/olmoe) | Allen Institute for AI, with Contextual AI | September 3, 2024 | 6.9B total; ~1.3B active per token | Apache 2.0 | Yes |
| [Qwen2.5-Coder](https://aiwiki.ai/wiki/qwen2_5_coder) | Qwen team, Alibaba | September 2024 | Six sizes, 0.5B to 32B (flagship 32.5B) | Apache 2.0 | Yes |
| [Qwen2-Math](https://aiwiki.ai/wiki/qwen2_math) | Qwen team, Alibaba | August 8, 2024 | 1.5B, 7B, and 72B (base and Instruct) | Apache 2.0 | Yes |
| [Llama 3.1](https://aiwiki.ai/wiki/llama_3_1) | Meta | July 23, 2024 | 8B, 70B, and 405B | Llama Community License | With conditions |
| [Mistral NeMo](https://aiwiki.ai/wiki/mistral_nemo) | Mistral AI, with NVIDIA | July 18, 2024 | 12B | Apache 2.0 | Yes |
| [SmolLM](https://aiwiki.ai/wiki/smollm) | Hugging Face | July 16, 2024 | 135M, 360M, and 1.7B variants | Apache 2.0 | Yes |
| [Gemma 2](https://aiwiki.ai/wiki/gemma_2) | Google DeepMind | June 27, 2024 | 2.6B, 9B, 27B | Gemma Terms of Use | With conditions |
| [LLM Compiler (Meta)](https://aiwiki.ai/wiki/llm_compiler) | Meta AI | June 27, 2024 | 7B and 13B (base and FTD variants) | Meta LLM Compiler License Agreement | With conditions |
| [Qwen2](https://aiwiki.ai/wiki/qwen2) | Alibaba Cloud | June 6, 2024 | 0.5B, 1.5B, 7B, 57B-A14B (14B active), 72B | Apache 2.0 | Yes |
| [Codestral](https://aiwiki.ai/wiki/codestral) (part of the family) | Mistral AI | May 29, 2024 | 22.2B (22B v0.1); 7.3B (Mamba); later versions not disclosed | Mistral Non-Production License | No |
| [DeepSeek-Prover](https://aiwiki.ai/wiki/deepseek_prover) | DeepSeek | May 23, 2024 | 7B; 671B MoE (V2) | DeepSeek Model License | With conditions |
| [DeepSeek-V2](https://aiwiki.ai/wiki/deepseek_v2) | DeepSeek | May 2024 | 236B total; ~21B active per token | DeepSeek Model License | With conditions |
| [IBM Granite](https://aiwiki.ai/wiki/ibm_granite) | IBM | May 2024 | ~350M to 32B MoE (H-Small, ~9B active) | Apache 2.0 | Yes |
| [Snowflake Arctic](https://aiwiki.ai/wiki/snowflake_arctic) | Snowflake AI Research | April 24, 2024 | 480B total; 17B active per token | Apache 2.0 | Yes |
| [Phi-3](https://aiwiki.ai/wiki/phi_3) | Microsoft | April 23, 2024 | 3.8B (Mini) to 14B (Medium) | MIT | Yes |
| [Llama 3](https://aiwiki.ai/wiki/llama_3) | Meta | April 18, 2024 | 8B and 70B (base and Instruct variants) | Llama Community License | With conditions |
| [Mixtral 8x22B](https://aiwiki.ai/wiki/mixtral_8x22b) | Mistral AI | April 17, 2024 | ~141B total; ~39B active per token | Apache 2.0 | Yes |
| [RecurrentGemma](https://aiwiki.ai/wiki/recurrentgemma) | Google DeepMind | April 11, 2024 | 2.68B and 8.58B total | Gemma Terms of Use | With conditions |
| [CodeGemma](https://aiwiki.ai/wiki/codegemma) | Google | April 9, 2024 | 2B to 7B | Gemma Terms of Use | With conditions |
| [Command R+](https://aiwiki.ai/wiki/command_r_plus) | Cohere | April 4, 2024 | 104 billion | CC BY-NC 4.0 | No |
| [Reka Edge](https://aiwiki.ai/wiki/reka_edge) (part of the family) | Reka AI | April 2024 | 7 billion | Reka Edge 2603 License (Business Source License 1.1) | With conditions |
| [Jamba](https://aiwiki.ai/wiki/jamba) | AI21 Labs | March 28, 2024 | 52B total; 12B active (Jamba 1.0) | Apache 2.0 | Yes |
| [DBRX](https://aiwiki.ai/wiki/dbrx) | Databricks (Mosaic AI) | March 27, 2024 | 132B total; 36B active per token | Databricks Open Model License | With conditions |
| [Command R](https://aiwiki.ai/wiki/command_r) | Cohere | March 11, 2024 | 35B (Command R); family spans 7B to 104B | CC BY-NC 4.0 | No |
| [BitNet b1.58](https://aiwiki.ai/wiki/bitnet_b1_58) | Microsoft Research | February 27, 2024 | 700M to 3B (original); 2B (BitNet b1.58 2B4T) | MIT | Yes |
| [Mistral Large](https://aiwiki.ai/wiki/mistral_large) (part of the family) | Mistral AI | February 26, 2024 | 123B (Large 2/2.1); 675B total, 41B active (Large 3) | Mistral Ecosystem Open-Weight License | With conditions |
| [Nemotron-4](https://aiwiki.ai/wiki/nemotron_4) (part of the family) | NVIDIA | February 26, 2024 | 15B; 340B (Base, Instruct, and Reward checkpoints) | NVIDIA Open Model License | With conditions |
| [Gemma](https://aiwiki.ai/wiki/gemma) | Google DeepMind | February 21, 2024 | 270 million to 31 billion across generations | Apache 2.0 | Yes |
| [DeepSeekMath](https://aiwiki.ai/wiki/deepseek_math) | DeepSeek | February 5, 2024 | 7 billion | DeepSeek Model License | With conditions |
| [MiniCPM](https://aiwiki.ai/wiki/minicpm) | OpenBMB | February 1, 2024 | 0.43B (MiniCPM4-0.5B) to 13.6B | Apache 2.0 | Yes |
| [OLMo](https://aiwiki.ai/wiki/olmo) | Allen Institute for AI (Ai2) | February 1, 2024 | 1B, 7B, 13B, 32B dense; OLMoE 1B active / 7B total | Apache 2.0 | Yes |
| [Aya (Expanse / Vision)](https://aiwiki.ai/wiki/aya_expanse) | Cohere Labs (Cohere) | February 2024 | 8B and 32B (Aya Expanse); 13B (Aya 101) | CC BY-NC 4.0 | No |
| [Reka Flash](https://aiwiki.ai/wiki/reka_flash) (part of the family) | Reka AI | February 2024 | 21 billion | Apache 2.0 | Yes |
| [GLM-4](https://aiwiki.ai/wiki/glm_4) (part of the family) | Zhipu AI | January 16, 2024 | Not stated | MIT | Yes |
| [Minitron](https://aiwiki.ai/wiki/minitron) | NVIDIA | 2024 | About 4B to 8B | NVIDIA Open Model License | With conditions |

### 2023 and earlier

| Model | Developer | Released | Parameters | Weights license | Commercial use |
|---|---|---|---|---|---|
| [Mixtral](https://aiwiki.ai/wiki/mixtral) | Mistral AI | December 11, 2023 | 8x7B: 46.7B total, 12.9B active | Apache 2.0 | Yes |
| [Mamba](https://aiwiki.ai/wiki/mamba) | Albert Gu and Tri Dao | December 1, 2023 | 130M to 2.8B (pretrained language models) | Apache 2.0 | Yes |
| [DeepSeek LLM](https://aiwiki.ai/wiki/deepseek_llm) | DeepSeek | November 29, 2023 | 7B and 67B | DeepSeek Model License | With conditions |
| [Grok](https://aiwiki.ai/wiki/grok) (part of the family) | xAI | November 4, 2023 | 314B, ~25% active (Grok-1); ~2.7T estimated (Grok-3) | Grok 2 Community License | With conditions |
| [DeepSeek-Coder](https://aiwiki.ai/wiki/deepseek_coder) | DeepSeek | November 2, 2023 | 1.3B, 6.7B, 33B dense (V1) | DeepSeek Model License | With conditions |
| [Yi (language model)](https://aiwiki.ai/wiki/yi) | 01.AI | November 2, 2023 | 6B and 34B base (family spans 1.5B to 34B) | Apache 2.0 | Yes |
| [Nemotron](https://aiwiki.ai/wiki/nemotron) | NVIDIA | November 2023 | 8B (2023) to ~550B total MoE (Nemotron 3 Ultra, 2026) | OpenMDW 1.1 | Yes |
| [BitNet](https://aiwiki.ai/wiki/bitnet) | Microsoft Research | October 17, 2023 | Not stated | MIT | Yes |
| [SEDD (Score Entropy Discrete Diffusion)](https://aiwiki.ai/wiki/sedd) | Stanford University (Aaron Lou, Chenlin Meng, Stefano Ermon) | October 2023 | Not stated | License not stated | Not stated |
| [Mistral 7B](https://aiwiki.ai/wiki/mistral_7b) | Mistral AI | September 27, 2023 | ~7.24 billion (announced as 7.3B) | Apache 2.0 | Yes |
| [Hunyuan](https://aiwiki.ai/wiki/hunyuan) (part of the family) | Tencent | September 7, 2023 | 389B total, 52B active (Hunyuan-Large) | Apache 2.0 | Yes |
| [Code Llama](https://aiwiki.ai/wiki/code_llama) | Meta AI | August 24, 2023 | 7B, 13B, 34B, 70B (base, Python, and Instruct variants) | Llama Community License | With conditions |
| [Llama 2](https://aiwiki.ai/wiki/llama_2) | Meta | July 18, 2023 | 7B, 13B, 70B (34B trained but withheld) | Llama Community License | With conditions |
| [InternLM](https://aiwiki.ai/wiki/internlm) | Shanghai AI Laboratory | July 6, 2023 | 1.8B to 20B across generations | Apache 2.0 | Yes |
| [Aquila (language model)](https://aiwiki.ai/wiki/aquila) | Beijing Academy of Artificial Intelligence (BAAI) | June 2023 | 7B, 34B, and experimental 70B | BAAI Aquila Model License Agreement | With conditions |
| [Phi (language model)](https://aiwiki.ai/wiki/phi) | Microsoft Research | June 2023 | 1.3B (Phi-1) to 15B (Phi-4-reasoning-vision) | MIT | Yes |
| [StarCoder](https://aiwiki.ai/wiki/starcoder) | BigCode project (Hugging Face and ServiceNow) | May 4, 2023 | 15.5B (StarCoder); 3B, 7B, 15B (StarCoder2) | OpenRAIL | With conditions |
| [Falcon (language model)](https://aiwiki.ai/wiki/falcon) | Technology Innovation Institute (TII) | May 2023 | 90M to 180B across generations | TII Falcon License | With conditions |
| [WizardLM](https://aiwiki.ai/wiki/wizardlm) | Microsoft | April 24, 2023 | Not stated | Llama Community License | With conditions |
| [Qwen](https://aiwiki.ai/wiki/qwen) (part of the family) | Alibaba Cloud (Qwen Team) | April 11, 2023 | Roughly 0.5B to over 1T across variants | Qwen Community License 1.0 | With conditions |
| [SenseNova](https://aiwiki.ai/wiki/sensenova) (part of the family) | SenseTime | April 10, 2023 | 2B to 27B (open-weight models); over 600B (SenseNova V6) | Apache 2.0 | Yes |
| [Vicuna (language model)](https://aiwiki.ai/wiki/vicuna) | LMSYS | March 30, 2023 | 7B, 13B, 33B | Llama Community License | With conditions |
| [ChatGLM](https://aiwiki.ai/wiki/chatglm) | Zhipu AI with Tsinghua University KEG lab | March 14, 2023 | ~6.2B (ChatGLM-6B); ~6B per later generation | ChatGLM Model License | With conditions |
| [CodeGeeX](https://aiwiki.ai/wiki/codegeex) | Tsinghua University (KEG) and Zhipu AI | March 2023 | 6B to 13B | CodeGeeX4 License | With conditions |
| [LLaMA](https://aiwiki.ai/wiki/llama) | Meta AI | February 24, 2023 | Up to 65B (LLaMA 1) | Llama Community License | With conditions |
| [SantaCoder](https://aiwiki.ai/wiki/santacoder) | BigCode (Hugging Face and ServiceNow Research) | January 9, 2023 | 1.1B | OpenRAIL | With conditions |
| [Atlas (language model)](https://aiwiki.ai/wiki/atlas_language_model) | Meta AI (FAIR) | August 2022 | Reader 220M to 11B, plus ~110M retriever | CC BY-NC 4.0 | No |
| [GLM-130B](https://aiwiki.ai/wiki/glm_130b) | Tsinghua KEG and Zhipu AI | August 2022 | 130 billion (dense) | GLM-130B License | No |
| [BLOOM (language model)](https://aiwiki.ai/wiki/bloom) | BigScience project (coordinated by Hugging Face) | July 12, 2022 | 176 billion (smaller variants 560M to 7.1B) | OpenRAIL | With conditions |
| [No Language Left Behind (NLLB)](https://aiwiki.ai/wiki/no_language_left_behind) | Meta AI | July 2022 | 600M to 54.5B | CC BY-NC 4.0 | No |
| [OPT (Open Pre-trained Transformer)](https://aiwiki.ai/wiki/opt) | Meta AI | May 3, 2022 | 125M to 175B (9 sizes) | OPT-175B License Agreement | No |
| [EXAONE](https://aiwiki.ai/wiki/exaone) (part of the family) | LG AI Research (LG Corporation) | December 14, 2021 | 1.2B to 32B dense | Apache 2.0 | Yes |
| [Sentence-transformers/all-MiniLM-L6-v2 model](https://aiwiki.ai/wiki/sentence-transformers_all-minilm-l6-v2_model) | Sentence Transformers project (2021 Hugging Face JAX/Flax community sprint) | August 2021 | ~22.7 million | Apache 2.0 | Yes |
| [Sentence-transformers/all-mpnet-base-v2 model](https://aiwiki.ai/wiki/sentence-transformers_all-mpnet-base-v2_model) | Sentence Transformers project (2021 Hugging Face Flax/JAX community sprint) | August 2021 | ~110 million | Apache 2.0 | Yes |
| [SPLADE](https://aiwiki.ai/wiki/splade) | Naver Labs Europe | July 12, 2021 | Not stated | CC BY-NC-SA 4.0 | No |
| [GPT-J](https://aiwiki.ai/wiki/gpt_j) | EleutherAI | June 9, 2021 | 6,053,381,344 (~6.05B) | Apache 2.0 | Yes |
| [GLM](https://aiwiki.ai/wiki/glm) (part of the family) | Tsinghua University and Zhipu AI (Z.ai) | June 2021 | 110M to 744B total (40B active, GLM-5) | MIT | Yes |
| [Huawei PanGu](https://aiwiki.ai/wiki/pangu) (part of the family) | Huawei (Huawei Cloud and Noah's Ark Lab) | April 2021 | 1.3B to 1.085T across the family | openPangu Model License Agreement 1.0 | With conditions |
| [Switch Transformer](https://aiwiki.ai/wiki/switch_transformer) | Google (Google Brain) | January 11, 2021 | Up to 1.571 trillion (Switch-C, 2,048 experts) | Apache 2.0 | Yes |
| [mT5](https://aiwiki.ai/wiki/mt5) | Google Research | October 22, 2020 | 300M to 13B (five sizes) | Apache 2.0 | Yes |
| [BlenderBot](https://aiwiki.ai/wiki/blenderbot) | Meta AI (FAIR) | April 29, 2020 | Up to 175B (BlenderBot 3) | MIT | Yes |
| [ColBERT](https://aiwiki.ai/wiki/colbert) | Stanford University (Omar Khattab, Matei Zaharia) | April 27, 2020 | 110M (ColBERTv2) | MIT | Yes |
| [Longformer](https://aiwiki.ai/wiki/longformer) | Allen Institute for AI | April 10, 2020 | Not stated | Apache 2.0 | Yes |
| [ELECTRA](https://aiwiki.ai/wiki/electra) | Stanford University and Google Brain | March 23, 2020 | 14M to 335M | Apache 2.0 | Yes |
| [DeBERTa](https://aiwiki.ai/wiki/deberta) | Microsoft Research | 2020 | 22M backbone (v3-XSmall) to 1.5B (v2-XXLarge) | MIT | Yes |
| [OPUS-MT](https://aiwiki.ai/wiki/opus-mt) | Language Technology Research Group, University of Helsinki | 2020 | Not stated | CC BY 4.0 | Yes |
| [LayoutLM](https://aiwiki.ai/wiki/layoutlm) | Microsoft Research | December 31, 2019 | ~113M (BASE) to ~343M (LARGE), v1 | MIT | Yes |
| [Sentence-BERT (SBERT)](https://aiwiki.ai/wiki/sentence-bert) | Nils Reimers and Iryna Gurevych, UKP Lab, TU Darmstadt | November 2019 | Not stated | Apache 2.0 | Yes |
| [XLM-RoBERTa](https://aiwiki.ai/wiki/xlm-roberta) | Facebook AI Research (Meta AI) | November 2019 | ~270M (Base); ~550M (Large); XL ~3.5B; XXL ~10.7B | MIT | Yes |
| [T5 (language model)](https://aiwiki.ai/wiki/t5) | Google | October 23, 2019 | 60 million to 11 billion (five sizes) | Apache 2.0 | Yes |
| [BART (language model)](https://aiwiki.ai/wiki/bart) | Facebook AI Research | October 2019 | ~140M (base); ~406M (large) | Apache 2.0 | Yes |
| [DistilBERT](https://aiwiki.ai/wiki/distilbert) | Hugging Face | October 2019 | 66 million | Apache 2.0 | Yes |
| [ALBERT](https://aiwiki.ai/wiki/albert) | Google Research and Toyota Technological Institute at Chicago | September 26, 2019 | 12M to 235M | Apache 2.0 | Yes |
| [RoBERTa](https://aiwiki.ai/wiki/roberta) | Facebook AI Research; University of Washington | July 26, 2019 | 125M (base), 355M (large) | MIT | Yes |
| [XLNet](https://aiwiki.ai/wiki/xlnet) | Carnegie Mellon University and Google Brain | June 2019 | 110M (Base), 340M (Large) | Apache 2.0 | Yes |
| [Baidu ERNIE](https://aiwiki.ai/wiki/baidu_ernie) (part of the family) | Baidu, Inc. | April 19, 2019 | Up to 424B total (47B active) in open-source ERNIE 4.5 | ERNIE 5.5 Flash Base Model License Agreement v1.0 | With conditions |
| [GPT-2](https://aiwiki.ai/wiki/gpt-2) | OpenAI | February 14, 2019 | 1.5B (XL); also 124M, 355M, 774M variants | Modified MIT | Yes |
| [CamemBERT](https://aiwiki.ai/wiki/camembert) | Inria (Almanach), Facebook AI Research, Sorbonne Universite | 2019 | 110M (base); 335M (large) | MIT | Yes |
| [SciBERT](https://aiwiki.ai/wiki/scibert) | Allen Institute for AI (AI2) | 2019 | ~110M | Apache 2.0 | Yes |
| [BERT](https://aiwiki.ai/wiki/bert) | Google AI Language | October 11, 2018 | 110M (Base); 340M (Large) | Apache 2.0 | Yes |
| [Bert-base-uncased model](https://aiwiki.ai/wiki/bert-base-uncased_model) | Google AI Language | October 2018 | ~110 million | Apache 2.0 | Yes |
| [GPT](https://aiwiki.ai/wiki/gpt) (part of the family) | OpenAI | June 11, 2018 | 117M (GPT-1) to 175B (GPT-3); undisclosed from GPT-4 onward | Modified MIT | Yes |
| [GPT-1](https://aiwiki.ai/wiki/gpt-1) | OpenAI | June 11, 2018 | 117 million | MIT | Yes |
| [ELMo (Embeddings from Language Models)](https://aiwiki.ai/wiki/elmo) | Allen Institute for AI and University of Washington | February 2018 | ~93.6 million | Apache 2.0 | Yes |

### Undated

| Model | Developer | Released | Parameters | Weights license | Commercial use |
|---|---|---|---|---|---|
| [Ling-3.1-flash](https://aiwiki.ai/wiki/ling_3_1_flash) | InclusionAI (Ant Group) | Not stated | About 560B total, 25B active (publisher's count) | MIT | Yes |
| [Nemotron-Labs-TwoTower](https://aiwiki.ai/wiki/nemotron_labs_twotower) | NVIDIA | Not stated | About 60B total (two 30B towers) | NVIDIA Open Model License | With conditions |


## Image models

| Model | Developer | Released | Parameters | Weights license | Commercial use |
|---|---|---|---|---|---|
| [Qwen-Image-2.1-Turbo](https://aiwiki.ai/wiki/qwen_image_2_1_turbo) | Qwen team (Alibaba) | October 9, 2026 | 7B (visual generation component) | Qwen Research License | No |
| [Ming-Image-0.1-Design](https://aiwiki.ai/wiki/ming_image_0_1_design) | inclusionAI (Ant Group) | September 22, 2026 | 6B diffusion transformer per model (developer figure) | MIT | Yes |
| [Qwen-Image-2.1](https://aiwiki.ai/wiki/qwen_image_2_1) | Qwen team, Alibaba | September 20, 2026 | 7B diffusion transformer; 8B Qwen3-VL text encoder | Qwen Research License | No |
| [FLUX.2](https://aiwiki.ai/wiki/flux_2) (part of the family) | Black Forest Labs | November 25, 2025 | ~32B (dev, pro, flex, max); klein 4B and 9B | FLUX Non-Commercial License | No |
| [Hunyuan Image 3.0](https://aiwiki.ai/wiki/hunyuan_image_3) | Tencent | September 28, 2025 | ~80B total; ~13B active per token; 64 experts | Tencent Hunyuan Community License | With conditions |
| [HiDream](https://aiwiki.ai/wiki/hidream) | HiDream.ai | April 2025 | About 17 billion | MIT | Yes |
| [Stable Diffusion 3.5](https://aiwiki.ai/wiki/stable_diffusion_3_5) | Stability AI | October 22, 2024 | 8.1B (Large); ~2.5B (Medium) | Stability AI Community License | With conditions |
| [Flux (text-to-image model)](https://aiwiki.ai/wiki/flux) (part of the family) | Black Forest Labs | August 1, 2024 | 12B (FLUX.1); 32B (FLUX.2); 4B and 9B (FLUX.2 klein) | FLUX Non-Commercial License | No |
| [FLUX.1](https://aiwiki.ai/wiki/flux_1) (part of the family) | Black Forest Labs | August 1, 2024 | ~12B | FLUX Non-Commercial License | No |
| [Stable Diffusion 3](https://aiwiki.ai/wiki/stable_diffusion_3) (part of the family) | Stability AI | June 12, 2024 | 800M to 8.1B (SD 3.5 Large) | Stability AI Community License | With conditions |
| [Latent Consistency Models (LCM)](https://aiwiki.ai/wiki/latent_consistency_models) | Tsinghua University (IIIS) | October 6, 2023 | Not stated | OpenRAIL | With conditions |
| [SDXL (Stable Diffusion XL)](https://aiwiki.ai/wiki/sdxl) | Stability AI | July 26, 2023 | ~2.6B U-Net; ~6.6B total with refiner | OpenRAIL | With conditions |
| [Würstchen](https://aiwiki.ai/wiki/wuerstchen) | Pablo Pernias, Dominic Rampas, Mats L. Richter, Christopher J. Pal, Marc Aubreville | June 1, 2023 | Stage A 18M; Stage B ~1B; Stage C ~0.99B | MIT | Yes |
| [ControlNet](https://aiwiki.ai/wiki/controlnet) | Lvmin Zhang, Anyi Rao and Maneesh Agrawala (Stanford University) | February 10, 2023 | Not stated | OpenRAIL | With conditions |
| [Stable Diffusion](https://aiwiki.ai/wiki/stable_diffusion) | Stability AI, with CompVis (LMU Munich), Runway, LAION, and EleutherAI | August 22, 2022 | 860M U-Net (SD 1.x); ~3.5B (SDXL base); 8.1B (SD 3.5 Large) | Stability AI Community License | With conditions |
| [GLIDE (OpenAI)](https://aiwiki.ai/wiki/glide) (part of the family) | OpenAI | December 20, 2021 | 3.5B (full model); ~300M released filtered model | MIT | Yes |
| [StyleGAN](https://aiwiki.ai/wiki/stylegan) | NVIDIA Research | December 12, 2018 | Not stated | CC BY-NC 4.0 | No |
| [CycleGAN](https://aiwiki.ai/wiki/cyclegan) | Berkeley AI Research (BAIR) Lab, UC Berkeley | March 2017 | Not stated | BSD 3-Clause | Yes |

## Video models

| Model | Developer | Released | Parameters | Weights license | Commercial use |
|---|---|---|---|---|---|
| [Pika H3 Flex](https://aiwiki.ai/wiki/pika_h3_flex) | Pika (Mellis, Inc.) | October 29, 2026 | 33B base generator | MiniMax H3 Community License Agreement | With conditions |
| [SenseNova 6.8 Code2Video](https://aiwiki.ai/wiki/sensenova_6_8_code2video) | SenseTime (SenseNova team) | October 2026 | 27B (27,356,728,560) | Apache 2.0 | Yes |
| [MAGI-2 Preview](https://aiwiki.ai/wiki/magi_2) | Sand.ai | August 5, 2026 | About 114B total, about 6B active | Apache 2.0 | Yes |
| [MiniMax H3](https://aiwiki.ai/wiki/minimax_h3) | MiniMax | July 31, 2026 | 33B (H3-Omni-Transformer generator) | MiniMax H3 Community License Agreement | With conditions |
| [Wan 2.1-VACE](https://aiwiki.ai/wiki/wan_vace) | Alibaba (Tongyi Lab) | May 14, 2025 | 1.3B and 14B (two variants) | Apache 2.0 | Yes |
| [Wan 2.1](https://aiwiki.ai/wiki/wan_2_1) | Alibaba (Tongyi Wanxiang team) | February 25, 2025 | 1.3B and 14B variants | Apache 2.0 | Yes |
| [Step-Video](https://aiwiki.ai/wiki/step_video) | StepFun (Shanghai Jieyue Xingchen) | February 17, 2025 | 30B (DiT backbone) | MIT | Yes |
| [HunyuanVideo](https://aiwiki.ai/wiki/hunyuan_video) | Tencent | December 3, 2024 | 13B (1.0); 8.3B (1.5) | Tencent Hunyuan Community License | With conditions |
| [LTX-Video](https://aiwiki.ai/wiki/ltx_video) | Lightricks | November 21, 2024 | ~1.9B (base); 13B (LTXV-13B) | LTXV Open Weights License | With conditions |
| [Mochi 1](https://aiwiki.ai/wiki/mochi_1) | Genmo Inc. | October 22, 2024 | 10 billion (DiT) plus 362 million (VAE) | Apache 2.0 | Yes |
| [Hailuo AI](https://aiwiki.ai/wiki/hailuo) (part of the family) | MiniMax | September 2024 | Not stated | MiniMax H3 Community License Agreement | With conditions |
| [CogVideoX](https://aiwiki.ai/wiki/cogvideo) | Zhipu AI and Tsinghua University (THUDM) | August 6, 2024 | 2B or 5B (by variant) | CogVideoX License | With conditions |
| [Open-Sora](https://aiwiki.ai/wiki/open_sora) | HPC-AI Tech (Colossal-AI team) | March 17, 2024 | 11B (v2.0) | Apache 2.0 | Yes |
| [Pika (video generation)](https://aiwiki.ai/wiki/pika) (part of the family) | Pika Labs | November 28, 2023 | Not stated | MiniMax H3 Community License Agreement | With conditions |
| [Stable Video Diffusion](https://aiwiki.ai/wiki/stable_video_diffusion) | Stability AI | November 21, 2023 | Not stated | Stability AI Community License | With conditions |

## Speech and audio models

| Model | Developer | Released | Parameters | Weights license | Commercial use |
|---|---|---|---|---|---|
| [Nemotron 3.5 Transcribe](https://aiwiki.ai/wiki/nemotron_3_5_transcribe) | NVIDIA | October 28, 2026 | About 34B total, about 4.3B active | OpenMDW 1.1 | Yes |
| [Whistle (Cactus Compute)](https://aiwiki.ai/wiki/whistle_cactus_compute) | Cactus Compute | October 2, 2026 | Not stated | Apache 2.0 | Yes |
| [Cohere Transcribe](https://aiwiki.ai/wiki/cohere_transcribe) | Cohere and Cohere Labs | March 26, 2026 | Not stated | Apache 2.0 | Yes |
| [Omnilingual ASR](https://aiwiki.ai/wiki/omnilingual_asr) | Meta FAIR | November 10, 2025 | 0.32B to 7.81B | Apache 2.0 | Yes |
| [Voxtral](https://aiwiki.ai/wiki/voxtral) | Mistral AI | July 15, 2025 | Voxtral Small 24.3B; Voxtral Mini 4.7B | Apache 2.0 | Yes |
| [Sesame CSM](https://aiwiki.ai/wiki/sesame_csm) (part of the family) | Sesame AI | March 13, 2025 | 1B backbone + 100M audio decoder (CSM-1B) | Apache 2.0 | Yes |
| [GLM-4-Voice](https://aiwiki.ai/wiki/glm_4_voice) | Zhipu AI and Tsinghua University KEG | October 25, 2024 | 9B (GLM-4-9B backbone) | glm-4-voice License | With conditions |
| [SpiRit-LM](https://aiwiki.ai/wiki/spirit_lm) | Meta AI (FAIR) | October 18, 2024 | 7B (Base and Expressive versions) | FAIR Noncommercial Research License | No |
| [F5-TTS](https://aiwiki.ai/wiki/f5_tts) | Shanghai Jiao Tong University X-LANCE Lab, University of Cambridge, Geely | October 9, 2024 | 335.8 million | CC BY-NC 4.0 | No |
| [Qwen2-Audio](https://aiwiki.ai/wiki/qwen2_audio) | Qwen team, Alibaba Cloud | August 2024 | About 8.2B | Apache 2.0 | Yes |
| [Moshi](https://aiwiki.ai/wiki/moshi) | Kyutai | July 3, 2024 | 7.7B total (Helium backbone is 7B) | CC BY 4.0 | Yes |
| [CosyVoice](https://aiwiki.ai/wiki/cosyvoice) | Tongyi SpeechTeam, Alibaba Group (FunAudioLLM project) | July 2024 | 300M (v1) | Apache 2.0 | Yes |
| [NVIDIA Canary](https://aiwiki.ai/wiki/canary) | NVIDIA | April 2024 | 182M to 2.5B across variants | CC BY 4.0 | Yes |
| [NVIDIA Parakeet](https://aiwiki.ai/wiki/parakeet) | NVIDIA (first models with Suno.ai) | December 2023 | 110M to 1.1B | CC BY 4.0 | Yes |
| [XTTS (Coqui XTTS)](https://aiwiki.ai/wiki/xtts) | Coqui AI | September 30, 2023 | About 750M | Coqui Public Model License 1.0.0 | No |
| [SeamlessM4T](https://aiwiki.ai/wiki/seamless_m4t) | Meta AI | August 22, 2023 | 1.2B (Medium), 2.3B (Large and v2) | CC BY-NC 4.0 | No |
| [MusicGen](https://aiwiki.ai/wiki/musicgen) | Meta AI (FAIR) | June 8, 2023 | 300M (Small), 1.5B (Medium), 3.3B (Large) | CC BY-NC 4.0 | No |
| [Massively Multilingual Speech (MMS)](https://aiwiki.ai/wiki/massively_multilingual_speech) | Meta AI | May 22, 2023 | 317M (MMS-300M), 965M (MMS-1B) | CC BY-NC 4.0 | No |
| [EnCodec](https://aiwiki.ai/wiki/encodec) | Meta AI (FAIR) | October 24, 2022 | Not stated | MIT | Yes |
| [Whisper](https://aiwiki.ai/wiki/whisper) | OpenAI | September 21, 2022 | 39M (tiny) to 1,550M (large) | MIT | Yes |
| [HuBERT](https://aiwiki.ai/wiki/hubert) | Meta AI (then Facebook AI Research) | June 14, 2021 | About 95M (Base), 317M (Large), 964M (X-Large) | Apache 2.0 | Yes |
| [Wav2Vec 2.0](https://aiwiki.ai/wiki/wav2vec2) | Meta (Facebook AI Research) | June 20, 2020 | About 95M (BASE), about 317M (LARGE) | Apache 2.0 | Yes |
| [Jukebox (OpenAI)](https://aiwiki.ai/wiki/jukebox) | OpenAI | April 30, 2020 | About 1B (each upsampling prior) to about 5B | OpenAI Jukebox Noncommercial Use License | No |
| [Wav2Vec](https://aiwiki.ai/wiki/wav2vec) | Meta AI (formerly Facebook AI Research) | 2019 | About 34M (wav2vec 1.0) to 2B (XLS-R 2B) | Apache 2.0 | Yes |
| [Ming-3.0](https://aiwiki.ai/wiki/ming_3_0) | inclusionAI | Not stated | 7.9B total, about 1.3B active (tts-tiny) | MIT | Yes |

## Robotics and physical AI models

| Model | Developer | Released | Parameters | Weights license | Commercial use |
|---|---|---|---|---|---|
| [AW-0](https://aiwiki.ai/wiki/aw_0) | Researchers at UC Berkeley, NVIDIA and the University of Pennsylvania | October 21, 2026 | Not stated | BSD 2-Clause | Yes |
| [LingBot-VLA 2.0](https://aiwiki.ai/wiki/lingbot_vla_2_0) | Robbyant (Ant Group) | July 2026 | 6B | Apache 2.0 | Yes |
| [NVIDIA Alpamayo 2 Super](https://aiwiki.ai/wiki/nvidia_alpamayo_2_super) | NVIDIA | June 1, 2026 | 34 billion | OpenMDW 1.1 | Yes |
| [NVIDIA Cosmos 3](https://aiwiki.ai/wiki/nvidia_cosmos_3) | NVIDIA | June 1, 2026 | 4B (Edge), 16B (Nano), 64B (Super) | OpenMDW 1.1 | Yes |
| [NVIDIA Isaac GR00T N1.7](https://aiwiki.ai/wiki/groot_n1_7) | NVIDIA | April 17, 2026 | 3B | NVIDIA Open Model License | With conditions |
| [Hy-Embodied](https://aiwiki.ai/wiki/hy_embodied) | Tencent (Hunyuan team and Tencent Robotics X) | April 2026 | About 4B (VLA-0.5), 6.2B (RxBrain-1.0) | Apache 2.0 | Yes |
| [NVIDIA Isaac GR00T N1.6](https://aiwiki.ai/wiki/groot_n1_6) | NVIDIA (GEAR team) | December 15, 2025 | 3B | NVIDIA License (non-commercial) | No |
| [V-JEPA 2](https://aiwiki.ai/wiki/v_jepa_2) | Meta AI | June 11, 2025 | 1.2B (world-model configuration; encoder over 1B) | MIT | Yes |
| [SmolVLA](https://aiwiki.ai/wiki/smolvla) | Hugging Face (LeRobot team) | June 3, 2025 | ~450M total (~100M action expert) | Apache 2.0 | Yes |
| [Isaac GR00T N1.5](https://aiwiki.ai/wiki/groot_n1_5) | NVIDIA | May 2025 | ~3B (GR00T-N1.5-3B) | NVIDIA License (non-commercial) | No |
| [NVIDIA Cosmos Reason](https://aiwiki.ai/wiki/cosmos_reason) | NVIDIA | May 2025 | 7B (Cosmos-Reason1); 2B to 32B (Cosmos-Reason2) | NVIDIA Open Model License | With conditions |
| [π0.5](https://aiwiki.ai/wiki/pi_0_5) | Physical Intelligence | April 22, 2025 | ~3.3B (~3B PaliGemma backbone + ~300M action expert) | Gemma Terms of Use | With conditions |
| [NVIDIA Isaac GR00T N1](https://aiwiki.ai/wiki/groot_n1) | NVIDIA (GEAR research lab) | March 18, 2025 | ~2B (base N1) | NVIDIA License (non-commercial) | No |
| [NVIDIA COMPASS](https://aiwiki.ai/wiki/nvidia_compass) | NVIDIA | February 2025 | Not stated | NVIDIA Open Model License | With conditions |
| [NVIDIA Cosmos](https://aiwiki.ai/wiki/nvidia_cosmos) | NVIDIA | January 6, 2025 | 4B to 14B (Cosmos 1.0 models) | OpenMDW 1.1 | Yes |
| [Meta Motivo](https://aiwiki.ai/wiki/meta_motivo) | Meta AI (FAIR) | December 12, 2024 | ~24.5M (S variants); ~288M (metamotivo-M-1) | CC BY-NC 4.0 | No |
| [π0](https://aiwiki.ai/wiki/pi0) | Physical Intelligence | October 31, 2024 | ~3.3B total (3B VLM backbone + ~300M action expert) | Gemma Terms of Use | With conditions |
| [OpenVLA](https://aiwiki.ai/wiki/openvla) | Stanford, UC Berkeley, TRI, Google DeepMind, Physical Intelligence, MIT | June 2024 | 7B | Llama Community License | With conditions |
| [Octo (Robot Policy)](https://aiwiki.ai/wiki/octo) | Octo Model Team | Not stated | 27M (Small) and 93M (Base) | MIT | Yes |

## Science models

| Model | Developer | Released | Parameters | Weights license | Commercial use |
|---|---|---|---|---|---|
| [Hydrogen-1.0](https://aiwiki.ai/wiki/hydrogen_1_0) | Deep Principle | November 2, 2026 | About 753B (753,329,940,480 in the BF16 checkpoint) | MIT | Yes |
| [Ling-3.0-flash-Sante](https://aiwiki.ai/wiki/ling_3_0_flash_sante) | inclusionAI | September 4, 2026 | 124B total, 5.1B active per token | MIT | Yes |
| [NVIDIA STRATA](https://aiwiki.ai/wiki/nvidia_strata) | NVIDIA, with U.S. national laboratories | June 30, 2026 | 331.7M | OpenMDW 1.1 | Yes |
| [rBio (CZI)](https://aiwiki.ai/wiki/czi_rbio) | Chan Zuckerberg Initiative (CZI) | August 21, 2025 | 3 billion (base model) | Qwen Research License | No |
| [UMA (Universal Model for Atoms)](https://aiwiki.ai/wiki/uma_meta) | FAIR Chemistry team, Meta AI | June 30, 2025 | UMA-S ~150M, UMA-M ~1.4B, UMA-L ~700M (total) | FAIR Chemistry License v1 | With conditions |
| [AlphaGenome](https://aiwiki.ai/wiki/alphagenome) | Google DeepMind | June 25, 2025 | Not stated | AlphaGenome Model Terms | No |
| [State (Arc Institute virtual cell model)](https://aiwiki.ai/wiki/arc_state_model) | Arc Institute | June 23, 2025 | SE-600M checkpoint about 0.7B | Arc Research Institute State Model Non-Commercial License | No |
| [Boltz-2](https://aiwiki.ai/wiki/boltz_2) | MIT (Jameel Clinic, CSAIL) and Recursion | June 2025 | Not stated | MIT | Yes |
| [MedGemma](https://aiwiki.ai/wiki/medgemma) | Google | May 20, 2025 | 4B and 27B | Health AI Developer Foundations Terms of Use | With conditions |
| [TxGemma](https://aiwiki.ai/wiki/txgemma) | Google | March 25, 2025 | 2B, 9B, 27B | Health AI Developer Foundations Terms of Use | With conditions |
| [Evo 2 (genomic model)](https://aiwiki.ai/wiki/evo_2) | Arc Institute, NVIDIA, Stanford, and collaborators | February 19, 2025 | About 40 billion (largest); also 7B and 1B | Apache 2.0 | Yes |
| [MatterGen](https://aiwiki.ai/wiki/mattergen) | Microsoft Research, AI for Science | January 16, 2025 | Not stated | MIT | Yes |
| [GenCast](https://aiwiki.ai/wiki/gencast) | Google DeepMind | December 4, 2024 | Not stated | CC BY 4.0 | Yes |
| [Boltz](https://aiwiki.ai/wiki/boltz) | MIT (CSAIL and Jameel Clinic) | November 1, 2024 | Not stated | MIT | Yes |
| [Chai-1](https://aiwiki.ai/wiki/chai_1) | Chai Discovery | September 9, 2024 | Not stated | Apache 2.0 | Yes |
| [ESM3](https://aiwiki.ai/wiki/esm3) (part of the family) | EvolutionaryScale | June 25, 2024 | 98B flagship; also 1.4B and 7B | MIT | Yes |
| [AlphaFold 3](https://aiwiki.ai/wiki/alphafold_3) | Google DeepMind and Isomorphic Labs | May 8, 2024 | Not stated | AlphaFold 3 Model Parameters Terms of Use | No |
| [NVIDIA Earth-2](https://aiwiki.ai/wiki/nvidia_earth_2) | NVIDIA | March 18, 2024 | Not stated | NVIDIA Open Model License | With conditions |
| [AlphaGeometry](https://aiwiki.ai/wiki/alphageometry) (part of the family) | Google DeepMind | January 17, 2024 | Up to 3.3B (AlphaGeometry 2 language model) | CC BY 4.0 | Yes |
| [GraphCast](https://aiwiki.ai/wiki/graphcast) | Google DeepMind | November 14, 2023 | 36.7M | CC BY 4.0 | Yes |
| [RFdiffusion](https://aiwiki.ai/wiki/rfdiffusion) | Baker Lab, UW Institute for Protein Design | March 30, 2023 | Not stated | BSD 3-Clause | Yes |
| [Galactica (language model)](https://aiwiki.ai/wiki/galactica) | Meta AI (Papers with Code team) | November 15, 2022 | Five sizes, 125M to 120B | CC BY-NC 4.0 | No |
| [ESMFold](https://aiwiki.ai/wiki/esmfold) | Meta AI (FAIR Protein Team) | November 1, 2022 | Production checkpoints use 3B ESM-2 stem | MIT | Yes |
| [BioGPT](https://aiwiki.ai/wiki/biogpt) | Microsoft Research | October 2022 | 347M (base); 1.5B (BioGPT-Large) | MIT | Yes |
| [OpenFold](https://aiwiki.ai/wiki/openfold) | AlQuraishi lab (Columbia University) and the OpenFold Consortium | June 2022 | Not stated | CC BY 4.0 | Yes |
| [AlphaFold-Multimer](https://aiwiki.ai/wiki/alphafold_multimer) | Google DeepMind | October 4, 2021 | Not stated | CC BY 4.0 | Yes |
| [BioBERT](https://aiwiki.ai/wiki/biobert) | DMIS Lab, Korea University | January 25, 2019 | 110M (Base); 340M (Large) | Apache 2.0 | Yes |

## Vision and multimodal models

| Model | Developer | Released | Parameters | Weights license | Commercial use |
|---|---|---|---|---|---|
| [WeVisOne-VL](https://aiwiki.ai/wiki/wevisone_vl) | Tencent (WeChat Vision team) | October 22, 2026 | 2B and 4B variants, each with a 9.6M pixel decoder | Apache 2.0 | Yes |
| [Intern-Decision](https://aiwiki.ai/wiki/intern_decision) | InternLM team, Shanghai AI Laboratory | September 26, 2026 | 0.8B, 2B and 4B | Apache 2.0 | Yes |
| [WeMM-Embedding](https://aiwiki.ai/wiki/wemm_embedding) | Tencent (WeChat Vision team) | August 25, 2026 | Not stated | Apache 2.0 | Yes |
| [HY-World 2.0](https://aiwiki.ai/wiki/hy_world_2) | Tencent Hunyuan | April 16, 2026 | About 425M to 80B per component | Tencent Hunyuan Community License | With conditions |
| [Lyra 2.0](https://aiwiki.ai/wiki/lyra_2_0) | NVIDIA Spatial Intelligence Lab | April 15, 2026 | Not stated | NVIDIA Internal Scientific R&D Model License | No |
| [DeepSeek-OCR](https://aiwiki.ai/wiki/deepseek-ocr) | DeepSeek | October 20, 2025 | About 3B decoder | MIT | Yes |
| [DINOv3](https://aiwiki.ai/wiki/dinov3) | Meta AI (FAIR) | August 14, 2025 | 21M to 6.7B (ViT-7B/16 largest) | DINOv3 License | With conditions |
| [HunyuanWorld 1.0](https://aiwiki.ai/wiki/hunyuanworld) | Tencent (Hunyuan team) | July 26, 2025 | Not stated | Tencent Hunyuan Community License | With conditions |
| [Qwen2.5-VL](https://aiwiki.ai/wiki/qwen2_5_vl) | Alibaba Cloud (Qwen team) | January 26, 2025 | 3B, 7B, 32B, 72B | Apache 2.0 | Yes |
| [DeepSeek-VL2](https://aiwiki.ai/wiki/deepseek_vl2) | DeepSeek (DeepSeek-AI) | December 13, 2024 | 3B/16B/27B total; 1.0B/2.8B/4.5B active | DeepSeek Model License | With conditions |
| [DeepSeek Janus](https://aiwiki.ai/wiki/deepseek_janus) | DeepSeek | October 17, 2024 | 1.3B (Janus, JanusFlow); 1B and 7B (Janus-Pro) | DeepSeek Model License | With conditions |
| [Molmo](https://aiwiki.ai/wiki/molmo) | Allen Institute for AI (Ai2) | September 25, 2024 | 72B, 7B (x2), MolmoE-1B (~1.5B active, 7.2B total) | Apache 2.0 | Yes |
| [Qwen2-VL](https://aiwiki.ai/wiki/qwen2_vl) | Alibaba Cloud (Qwen team) | August 29, 2024 | 2B, 7B, 72B | Apache 2.0 | Yes |
| [Sapiens (computer vision)](https://aiwiki.ai/wiki/sapiens) | Meta (Reality Labs) | August 22, 2024 | 0.3B, 0.6B, 1B, 2B | CC BY-NC 4.0 | No |
| [SAM 2](https://aiwiki.ai/wiki/sam_2) | Meta AI (FAIR) | July 29, 2024 | Tiny 38.9M, Small 46M, Base+ 80.8M, Large 224.4M | Apache 2.0 | Yes |
| [Chameleon (Meta AI)](https://aiwiki.ai/wiki/chameleon) | Meta AI (FAIR) | June 18, 2024 | 7B and 34B | Chameleon Research License | No |
| [Florence-2](https://aiwiki.ai/wiki/florence_2) | Microsoft Research | June 16, 2024 | 232M (base); 771M (large) | MIT | Yes |
| [PaliGemma](https://aiwiki.ai/wiki/paligemma) | Google | May 14, 2024 | 3B (PaliGemma); 3B, 10B and 28B (PaliGemma 2) | Gemma Terms of Use | With conditions |
| [DeepSeek-VL](https://aiwiki.ai/wiki/deepseek_vl) | DeepSeek | March 11, 2024 | 1.3B and 7B | DeepSeek Model License | With conditions |
| [V-JEPA](https://aiwiki.ai/wiki/v_jepa) | Meta AI (FAIR) | February 15, 2024 | Up to ~630M (ViT-H/16 encoder) | CC BY-NC 4.0 | No |
| [MiniCPM-V](https://aiwiki.ai/wiki/minicpm_v) | OpenBMB (THUNLP and ModelBest Inc.) | February 1, 2024 | 1.3B to 9B, depending on version | Apache 2.0 | Yes |
| [InternVL](https://aiwiki.ai/wiki/internvl) | OpenGVLab (Shanghai AI Laboratory) and academic partners | December 21, 2023 | 0.94B to 240.7B (28B active in InternVL3.5-241B-A28B) | Apache 2.0 | Yes |
| [CogVLM](https://aiwiki.ai/wiki/cogvlm) | Zhipu AI and Tsinghua University (KEG) | October 5, 2023 | About 17B (about 10B vision, 7B language) | CogVLM License | With conditions |
| [Nougat (model)](https://aiwiki.ai/wiki/nougat) | Meta AI | August 25, 2023 | 250M (small); 350M (base) | CC BY-NC 4.0 | No |
| [Qwen-VL](https://aiwiki.ai/wiki/qwen_vl) (part of the family) | Alibaba Cloud (Qwen team) | August 22, 2023 | 9.6B | Qwen License (Tongyi Qianwen) | With conditions |
| [Hiera](https://aiwiki.ai/wiki/hiera) | Meta AI (FAIR) | June 1, 2023 | 28M to 673M | CC BY-NC 4.0 | No |
| [ImageBind](https://aiwiki.ai/wiki/imagebind) | Meta AI (FAIR) | May 9, 2023 | Not stated | CC BY-NC-SA 4.0 | No |
| [DINOv2](https://aiwiki.ai/wiki/dinov2) | Meta AI Research (FAIR) | April 14, 2023 | 21M to 1.1B (ViT-S/14, ViT-B/14, ViT-L/14, ViT-g/14) | Apache 2.0 | Yes |
| [Segment Anything Model and Dataset (SAM and SA-1B)](https://aiwiki.ai/wiki/segment_anything_model_and_dataset_sam_and_sa-1b) | Meta AI | April 5, 2023 | ~91M (ViT-B), ~308M (ViT-L), ~636M (ViT-H) | Apache 2.0 | Yes |
| [DINO (computer vision)](https://aiwiki.ai/wiki/dino_model) | Meta AI | April 2023 | 21M (ViT-S) to 6.7B (DINOv3 ViT-7B) | DINOv3 License | With conditions |
| [LLaVA (Large Language and Vision Assistant)](https://aiwiki.ai/wiki/llava) | University of Wisconsin-Madison, Microsoft Research and Columbia University | April 2023 | 7B and 13B backbones (LLaVA-1.5) | Apache 2.0 | Yes |
| [SigLIP](https://aiwiki.ai/wiki/siglip) | Google DeepMind | March 27, 2023 | 86M to 1B (SigLIP 2, ViT-B to ViT-g) | Apache 2.0 | Yes |
| [YOLOv8](https://aiwiki.ai/wiki/yolov8) | Ultralytics | January 10, 2023 | 3.2M (YOLOv8n) to 68.2M (YOLOv8x) for detection | GNU AGPL-3.0 | Yes |
| [ConvNeXt](https://aiwiki.ai/wiki/convnext) | Facebook AI Research (FAIR) and UC Berkeley | January 2023 | 3.7M (V2-Atto) to 660M (V2-Huge) | CC BY-NC 4.0 | No |
| [I-JEPA](https://aiwiki.ai/wiki/i_jepa) | Meta AI | January 2023 | About 632M (ViT-Huge) | CC BY-NC 4.0 | No |
| [MetaCLIP](https://aiwiki.ai/wiki/metaclip) | Meta AI | 2023 | Not stated | CC BY-NC 4.0 | No |
| [InternVideo](https://aiwiki.ai/wiki/internvideo) | OpenGVLab, Shanghai AI Laboratory; Nanjing University; SIAT, CAS | December 6, 2022 | Up to 6B (InternVideo2-6B) | Apache 2.0 | Yes |
| [Donut (Model)](https://aiwiki.ai/wiki/donut_model) | NAVER CLOVA | November 2021 | 143M (176M when English token embeddings are counted) | MIT | Yes |
| [CLIP (Contrastive Language-Image Pre-training)](https://aiwiki.ai/wiki/clip) | OpenAI | 2021 | Not stated | MIT | Yes |
| [Swin Transformer](https://aiwiki.ai/wiki/swin_transformer) | Microsoft Research Asia | 2021 | 29M (Swin-T) to 197M (Swin-L); about 3B (SwinV2-G) | MIT | Yes |
| [DeiT](https://aiwiki.ai/wiki/deit) | Facebook AI Research (FAIR) and Sorbonne University | December 23, 2020 | 5M to 87M | Apache 2.0 | Yes |
| [DETR](https://aiwiki.ai/wiki/detr) | Facebook AI Research (FAIR) | 2020 | 41M (ResNet-50) to 60M (ResNet-101) | Apache 2.0 | Yes |
| [EfficientNet](https://aiwiki.ai/wiki/efficientnet) | Google Brain (Mingxing Tan, Quoc V. Le) | 2019 | 5.3M (B0) to 66M (B7) | Apache 2.0 | Yes |
| [ResNet](https://aiwiki.ai/wiki/resnet) | Kaiming He, Xiangyu Zhang, Shaoqing Ren and Jian Sun | 2016 | 25.6M (ResNet-50, TorchVision count) | MIT | Yes |
| [YOLO (object detection)](https://aiwiki.ai/wiki/yolo) | Joseph Redmon and co-authors (v1 to v3); later versions by Ultralytics and other groups | June 2015 | Varies by version; 55.7M for YOLO26x | GNU AGPL-3.0 | Yes |
| [GoogLeNet (Inception v1)](https://aiwiki.ai/wiki/googlenet) | Google Research | September 17, 2014 | ~5M (~6.8M counting all layers incl. auxiliary heads) | Apache 2.0 | Yes |
| [VGGNet](https://aiwiki.ai/wiki/vggnet) | Visual Geometry Group, University of Oxford | September 4, 2014 | 138.4M (VGG-16); 143.7M (VGG-19) | CC BY 4.0 | Yes |
| [d3-turbo](https://aiwiki.ai/wiki/d3_turbo) | vLLM Semantic Router team | Not stated | 26.1B loaded | Apache 2.0 | Yes |

## Other models

| Model | Developer | Released | Parameters | Weights license | Commercial use |
|---|---|---|---|---|---|
| [Decision 2.0](https://aiwiki.ai/wiki/decision_2_0) | vLLM Semantic Router project | October 3, 2026 | 0.60B to 29.37B (six checkpoints, nominal sizes 0.6B to 27B) | Apache 2.0 | Yes |
| [Extropic Z1T](https://aiwiki.ai/wiki/extropic_z1t) | Extropic | September 4, 2026 | Not stated | License not stated | Not stated |
| [Frontis-MA1](https://aiwiki.ai/wiki/frontis_ma1) | FrontisAI (Horizon Research of Frontis.AI with Tsinghua University) | July 30, 2026 | 35B total, about 3B active (Frontis-MA1-35B) | CC BY-NC 4.0 | No |
| [NVIDIA Ising](https://aiwiki.ai/wiki/nvidia_ising) | NVIDIA | April 14, 2026 | 31B (Calibration 1.5); 35B total, 3B active (Calibration 1) | OpenMDW 1.1 | Yes |
| [Xiaomi MiMo-V2.5](https://aiwiki.ai/wiki/mimo_v2_5) | Xiaomi | April 2026 | 310B total, 15B active (standard) | MIT | Yes |
| [Qwen-Image](https://aiwiki.ai/wiki/qwen_image) | Qwen team, Alibaba (Tongyi Lab) | August 4, 2025 | ~20 billion (MMDiT backbone) | Apache 2.0 | Yes |
| [Tülu 3](https://aiwiki.ai/wiki/tulu_3) | Allen Institute for AI (Ai2) | November 21, 2024 | 8B, 70B, 405B | Llama Community License | With conditions |
| [Hunyuan 3D](https://aiwiki.ai/wiki/hunyuan_3d) (part of the family) | Tencent | November 2024 | Up to ~10B shape model (2.5); 3.3B shape + 2B paint in 2.1 | Tencent Hunyuan Community License | With conditions |
| [Show-o](https://aiwiki.ai/wiki/show_o) | Show Lab, National University of Singapore (with ByteDance co-authors) | August 21, 2024 | About 1.3B | Apache 2.0 | Yes |
| [ShieldGemma](https://aiwiki.ai/wiki/shieldgemma) | Google | July 31, 2024 | 2B, 9B, 27B (text); 4B (ShieldGemma 2) | Gemma Terms of Use | With conditions |
| [BGE (BAAI General Embedding)](https://aiwiki.ai/wiki/bge) | Beijing Academy of Artificial Intelligence (BAAI) | August 2, 2023 | 24M to 326M (v1.5 embedders) | MIT | Yes |
| [Shap-E](https://aiwiki.ai/wiki/shap_e) | OpenAI | May 3, 2023 | Not stated | MIT | Yes |
| [Alpaca (model)](https://aiwiki.ai/wiki/alpaca) | Stanford CRFM | March 13, 2023 | 7B (13B variant documented) | CC BY-NC 4.0 | No |
| [RWKV](https://aiwiki.ai/wiki/rwkv) | Bo Peng (BlinkDL) and the RWKV project, hosted by the Linux Foundation | August 2021 | 169M to 14B | Apache 2.0 | Yes |
| [AlphaFold](https://aiwiki.ai/wiki/alphafold) | Google DeepMind | July 15, 2021 | Not stated | AlphaFold 3 Model Parameters Terms of Use | No |
| [DistilGPT2](https://aiwiki.ai/wiki/distilgpt2_model) | Hugging Face | October 3, 2019 | 82M (81,912,576 learned parameters) | Apache 2.0 | Yes |

## See also

- [Open weights](https://aiwiki.ai/wiki/open_weights)
- [Open-source AI](https://aiwiki.ai/wiki/open_source_ai)
- [Open Source AI Definition](https://aiwiki.ai/wiki/open_source_ai_definition)
- [Open-weight license comparison](https://aiwiki.ai/wiki/open_weight_license_comparison)
- [OpenRAIL](https://aiwiki.ai/wiki/openrail)
- [OpenMDW](https://aiwiki.ai/wiki/openmdw)
- [Model Openness Framework](https://aiwiki.ai/wiki/model_openness_framework)
- [Openwashing](https://aiwiki.ai/wiki/openwashing)
- [Best local LLMs](https://aiwiki.ai/wiki/best_local_llms)
- [llama.cpp](https://aiwiki.ai/wiki/llama_cpp)
- [Ollama](https://aiwiki.ai/wiki/ollama)

## References

1. Open Source Initiative. "The Open Source AI Definition 1.0". https://opensource.org/ai/open-source-ai-definition
2. Apache Software Foundation. "Apache License, Version 2.0". January 2004. https://www.apache.org/licenses/LICENSE-2.0
3. Open Source Initiative. "The MIT License". https://opensource.org/license/mit
4. Creative Commons. "Attribution 4.0 International (CC BY 4.0)" (deed). https://creativecommons.org/licenses/by/4.0/
5. Creative Commons. "Attribution-NonCommercial 4.0 International (CC BY-NC 4.0)" (deed). https://creativecommons.org/licenses/by-nc/4.0/

