Tencent Hunyuan Hy3 (marketed internationally as Tencent Hy3) is an open-weights large language model published by Tencent. It was first released as Hy3-preview on April 23, 2026, officially released as Hy3 on July 6, 2026, and rolled out globally across Tencent products and cloud services on August 5, 2026.[10][12] It is a Mixture-of-Experts (MoE) reasoning model and agent model with roughly 295 billion total parameters and about 21 billion active per token, and it was the first major model to come out of a rebuild of Tencent's pre-training and reinforcement-learning systems that began in early 2026.[1][2][3][10] Tencent positions it as a cost-efficient open model aimed at practical work: complex reasoning, instruction following, coding, and multi-step agent tasks rather than chart-topping leaderboard runs.[2][4]

The "Hy" in the name is Tencent's shortened overseas brand for Hunyuan. In 2026 the company simplified its international model brand from "Hunyuan" to "Tencent HY," so "Hy3" reads as the third major generation of the Hunyuan family.[5] Tencent's August 2026 global announcement formalizes the branding as "Tencent Hy3 (Tencent Hy, formerly known as Tencent Hunyuan)".[12] The repositories and model cards still sit under the Tencent-Hunyuan organization, and Chinese-language coverage refers to the same release as the Hunyuan Hy3 preview.[1][3]

What it is

Hy3-preview is a text in, text out language model trained to do extended reasoning before answering. Tencent describes it as a fused fast-and-slow-thinking model, meaning a single model that can answer quickly on easy prompts and spend more compute reasoning step by step on hard ones, rather than shipping a separate "thinking" variant.[1][3] The preview shipped in two public checkpoints: an instruction-tuned model (tencent/Hy3-preview) and a pre-trained base model (tencent/Hy3-preview-Base).[6] The official July 2026 release added the finished instruction-tuned checkpoint (tencent/Hy3) and an FP8-quantized variant (tencent/Hy3-FP8).[11]

The word "preview" is deliberate. Tencent released the model as an early, still-improving snapshot of the rebuilt Hunyuan line rather than a finished flagship. Yao Shunyu, the company's chief AI scientist, described it as the first step in rebuilding the Hunyuan model line, and said Tencent is "continuously expanding the scale of our pre-training and reinforcement learning efforts to push the boundaries of model intelligence."[2][4] Yao previously worked at OpenAI and is known in the research community for the ReAct framework that interleaves reasoning and actions for agents, which fits the model's agent focus.[4]

Tencent's Hunyuan line and where Hy3 fits

Hunyuan is Tencent's in-house family of foundation models, and the lineup grew quickly across 2024 and 2025 before the Hy3 rebuild.

Model	Released	Type	Notes
Hunyuan Large	Nov 2024	MoE text	Early open MoE release, around 389B total / 52B active
Hunyuan 3D	Nov 2024	3D generation	Open-sourced 3D asset model
HunyuanVideo	Dec 2024	Text to video	Open video generation model
Hunyuan-T1	Mar 2025	Reasoning	Tencent's first deep-thinking model
Hunyuan-A13B	Jun 2025	MoE reasoning	80B total / 13B active, fine-grained MoE
Hunyuan 2.0 (Hy2)	Dec 2025	MoE	Prior generation before the rebuild
Hunyuan Hy3 preview	Apr 2026	MoE reasoning/agent	295B total / 21B active, first post-rebuild model
Hy3 (official release)	Jul 2026	MoE reasoning/agent	Same 295B / 21B architecture, scaled-up post-training, Apache 2.0

Hy3 is the successor to the Hunyuan 2.0 generation (referred to as Hy2 in some coverage). Tencent frames it as a clean break: starting in early 2026 the team rebuilt its pre-training and reinforcement-learning infrastructure with a stated focus on systematic capability, honest evaluation, and cost-effectiveness, then trained this first post-rebuild model in under three months.[2][3][4] Earlier Hunyuan models such as Hunyuan-A13B remain separate, smaller releases and are not the same model as Hy3.

Architecture

Hy3-preview uses a sparse MoE transformer. Of the roughly 295 billion total parameters, only about 21 billion are active for any given token, which is what keeps inference cheap relative to a dense model of similar quality.[1][6] The published configuration lists the following.

Property	Value
Total parameters	~295B
Active parameters per token	~21B
Multi-token prediction (MTP) layer	3.8B parameters, 1 layer
Transformer layers	80 (excluding the MTP layer)
Experts	192 routed experts, top-8 activated, plus shared experts
Attention	Grouped-query attention, 64 query heads over 8 key/value heads, head dim 128
Hidden size	4,096
Intermediate size	13,312
Vocabulary	120,832 tokens
Context length	256K tokens
Precision	BF16

Two design choices stand out. First, the model carries a dedicated multi-token prediction (MTP) layer of about 3.8 billion parameters that predicts more than one token at a time. This enables speculative decoding for faster generation, and it is wired into serving stacks such as vLLM.[6] Second, reporting on the architecture describes a differentiated expert-size design, where experts are not all the same width and tokens of varying difficulty can be routed to experts with different capacities, in contrast to uniform-expert MoE layers.[7] The 256K context window puts it in the same range as other 2026 frontier open models for long-document and long-agent-trajectory work.[1][6]

Reasoning and agent capabilities

The pitch for Hy3 is less about a single benchmark and more about being useful inside real agent loops. Tencent reports the model was deployed across its own products before the public release, including the Yuanbao assistant, the CodeBuddy and WorkBuddy coding agents, the ima note tool, Tencent Docs, and the game Peacekeeper Elite.[1][3]

The product numbers Tencent cites come from those deployments. The company says the rebuilt model cut overall reasoning efficiency cost by about 40 percent versus the previous generation, reduced time-to-first-token on CodeBuddy and WorkBuddy by 54 percent, and cut end-to-end response time by 47 percent, while reporting a 99.99 percent task success rate and support for agent workflows running up to 495 steps.[1][2] For Tencent Docs' AI slide-generation feature, it reports a 20 percent increase in generation success rate.[1] These are vendor-reported figures from internal products, so they are best read as Tencent's own measurements rather than independent results.

Benchmarks

Tencent has been openly skeptical of public leaderboards, saying it moved toward self-built tests, human review, and product beta testing because public benchmarks can be gamed.[4] That said, the model cards and coverage do report standard numbers. The instruction-tuned model's published scores include the following.

Benchmark	Hy3-preview (Instruct)
SWE-bench Verified (coding agent)	74.4%
Terminal-Bench 2.0 (terminal agent)	54.4%
BrowseComp (search agent)	67.1%
GPQA Diamond (graduate science QA)	87.2
Humanity's Last Exam (HLE)	~30

Sources: model card and reporting.[6][8][9]

On STEM and reasoning, Tencent highlights strong results on hard science and math tasks such as a FrontierScience-Olympiad set and IMOAnswerBench, plus what it calls excellent results on the Tsinghua Qiuzhen College math PhD qualifying exam (spring 2026) and the China High School Biology Olympiad (CHSBO 2025), though it does not publish single headline numbers for all of these.[1][6] The pre-trained base model posts results including roughly 95.4% on GSM8K (4-shot), 76.3% on MATH (4-shot), 87.4% on MMLU (5-shot), 65.8% on MMLU-Pro, and 96.0% on ARC-Challenge, benchmarked against the base models of Kimi-K2, DeepSeek-V3, and GLM-4.5.[6]

The headline coding result, about 74 on SWE-bench Verified, is the figure Tencent leans on, and it is competitive for an open model of this size. It still trails the strongest proprietary frontier systems of the period: coverage placing Hy3 next to closed models reports roughly 80.8% for Claude Opus 4.6 and 78.6% for GPT-5.4 on the same test, so Hy3 is closing the gap rather than leading it.[8] Against open competitors it is described as roughly competitive with GLM-5.[8]

Official release (July 6, 2026)

Tencent released the official Hy3 on July 6, 2026, dropping the preview label after roughly ten weeks of public testing.[10] The architecture is unchanged from the preview (295B total parameters, 21B active, one 3.8B-parameter MTP layer, 256K context), but Tencent says it scaled up post-training compute and improved the quality and diversity of post-training data, drawing on feedback from more than 50 of its own products.[10][11] The company reported the largest gains in productivity tasks: software development, office work, financial modeling, front-end design, and game production.[10][11]

The official model card frames the improvements in reliability terms rather than leaderboard scores. Tencent reports that tool-call and output-format stability reached production grade, with SWE-bench Verified accuracy varying within 4 percent across agent scaffoldings such as CodeBuddy, Cline, and KiloCode; that the hallucination rate in internal real-world evaluations dropped from 12.5 percent to 5.4 percent and commonsense error rates from 25.4 percent to 12.7 percent; and that the issue rate on internal multi-turn dialogue tests fell from 17.4 percent to 7.9 percent.[11] Instead of a single benchmark headline, Tencent cites a blind evaluation in which 270 experts scored models on tasks drawn from their own work: Hy3 scored 2.67 out of 4 against 2.51 for GLM-5.1, with the largest advantages in front-end development, data and storage, and CI/CD tasks.[11] All of these are Tencent's own evaluations.

At the release Tencent said daily token consumption had grown 20x since the preview went live, and that in internal evaluations inside WorkBuddy, its agentic workplace product, the task success rate rose from 72 percent to 90 percent while average task completion time fell 34 percent versus the previous Hy model.[10][12] The released model exposes a reasoning_effort parameter (no_think, low, high) that controls whether it answers directly or produces a deep chain of thought, implementing the fused fast-and-slow-thinking design as an API-level switch.[11] The weights were published the same day on Hugging Face, ModelScope, GitCode, and CNB under the Apache 2.0 license, alongside an FP8-quantized Hy3-FP8 variant.[10][11] The model was also listed on OpenRouter on July 6.[14]

Global availability (August 2026)

On August 5, 2026, Tencent announced broader international access to Hy3, positioning it as the flagship of the renamed Tencent Hy model family.[12][15] The model is available to users and enterprises through WorkBuddy, free of charge worldwide until August 31, 2026 (Pacific Time), as well as through the AI design studio Tencent Design Miora and Tencent Cloud TokenHub, a model-as-a-service platform, with API access for developers.[12] Tencent Cloud is also working with regional partners, the South Korean e-commerce platform provider Cafe24 and the Japanese enterprise AI platform company Metelix, to offer Hy3 through their gateway and platform services.[12]

In the announcement, Tencent said Hy3 recorded more than 68 times as many API calls as its previous-generation model and ranked first on OpenRouter's global LLM usage leaderboard within one week of its launch, and claimed the model "performs comparably to flagship models with two to five times as many parameters."[12] The South China Morning Post, covering the announcement, reported that as of August 5 the model ranked third on OpenRouter behind DeepSeek's V4 Flash and Xiaomi's MiMo-V2.5.[13] On OpenRouter, Hy3 is listed starting at US$0.1288 per million input tokens and US$0.5336 per million output tokens.[12][14] Beyond OpenRouter, Tencent lists Hy3 on third-party developer platforms including Hermes, Kilo, Cline, OpenClaw, OpenCode, and Cherry Studio, and the model has been available on Hugging Face and ModelScope since release day.[12]

"Hy3 marks a significant step in how Tencent Cloud International is helping organisations move beyond AI experimentation and put intelligence to work in practical, trusted and scalable ways," said Poshu Yeung, Senior Vice President of Tencent Cloud and Head of Tencent Cloud International, in the announcement.[12]

Open-weights release and license

Tencent open-sourced Hy3-preview's weights on the same day it announced the model. The checkpoints are available on Hugging Face at tencent/Hy3-preview and tencent/Hy3-preview-Base, on ModelScope, and on GitCode, with code and documentation in the Tencent-Hunyuan/Hy3-preview GitHub repository.[1][6] The model is also offered through Tencent Cloud's TokenHub API and was listed on OpenRouter with a two-week free-access window at launch.[2]

The preview weights were released under the Tencent Hy Community License Agreement, a custom license rather than a standard permissive one such as Apache 2.0 or MIT, which meant the preview checkpoints were downloadable and usable but subject to Tencent's own terms.[6] The license changed at the official release: the tencent/Hy3 model card and its bundled LICENSE file state that Hy3 is released under the Apache License 2.0, and Tencent's August 2026 global announcement describes the model as available "under the commercially permissive Apache 2.0 license," a standard open-source grant that permits free commercial use and modification.[11][12] The preview repositories on Hugging Face retain the older community license.[6][11]

Cost efficiency and context

Cost is central to the Hy3 story. With only about 21 billion active parameters and the MTP-based speculative decoding, the model is cheap to serve for its quality tier, which is also reflected in Tencent Cloud's launch pricing: TokenHub listed Hy3-preview at 1.2 yuan per million input tokens, 0.4 yuan per million for cached input, and 4 yuan per million output tokens, in the range of roughly $0.18, $0.06, and $0.59 per million tokens at the time.[2][4] That places it among the aggressively priced open frontier models of 2026 alongside the DeepSeek and Qwen families.

The broader context is Tencent's open-weights strategy. By shipping a capable agent-oriented model under a community license, integrating it across its own product surface, and competing largely on price and practical reliability, Tencent is following the same playbook that made other Chinese labs influential in the open-model ecosystem.[4][8] Hy3-preview's "preview" label and Yao Shunyu's framing of it as a first step suggest Tencent intends further releases on the rebuilt infrastructure, so the model is best understood as the opening move of a new Hunyuan generation rather than its final form.[2][4]

References

Tencent, "Tencent Unveils Hy3 preview; Model Enhances Agent Capabilities and Real-World Usability," tencent.com, April 2026. https://www.tencent.com/en-us/articles/2202320.html
implicator.ai, "Tencent ships Hy3-preview as 295B model with 40 percent efficiency gain," April 2026. https://www.implicator.ai/tencent-ships-hy3-preview-as-295b-model-with-40-percent-efficiency-gain/
Tencent Hunyuan, "Hy3-preview," GitHub (Tencent-Hunyuan/Hy3-preview). https://github.com/Tencent-Hunyuan/Hy3-preview
techi.com, "Tencent's Hy3 preview Marks the First Major Model to Emerge From Its AI Rebuild." https://www.techi.com/tencent-hy3-preview-hunyuan-ai-model/
aibase.com, "Tencent's Overseas Large Model Brand Rebranding: Hunyuan Simplified to Tencent HY." https://www.aibase.com/news/23539
Tencent, "tencent/Hy3-preview," Hugging Face. https://huggingface.co/tencent/Hy3-preview
techjacksolutions.com, "Tencent Open-Sources Hy3-preview: 295B MoE Model with 21B Active Parameters Built for STEM Reasoning." https://techjacksolutions.com/ai-brief/tencent-open-sources-hy3-preview-295b-moe-model-with-21b-act/
toolworthy.ai, "Tencent Hy3 Preview (2026): Open 295B MoE, 74.4 SWE-Bench." https://www.toolworthy.ai/tool/tencent-hy-hy3-preview
artificialanalysis.ai, "DeepSeek V4 Pro vs Hy3-preview: Model Comparison." https://artificialanalysis.ai/models/comparisons/deepseek-v4-pro-vs-hy3
Tencent (official account), "腾讯混元 Hy3，正式发布" (Tencent Hunyuan Hy3 officially released), news.qq.com, July 6, 2026. https://news.qq.com/rain/a/20260706A07GV200
Tencent, "tencent/Hy3," Hugging Face model card, July 2026. https://huggingface.co/tencent/Hy3
Tencent, "Tencent Hy3 Now Available Globally, Extending Practical AI Across Products, Workflows and Cloud Services," PRNewswire, August 5, 2026. https://www.prnewswire.com/news-releases/tencent-hy3-now-available-globally-extending-practical-ai-across-products-workflows-and-cloud-services-302843373.html
South China Morning Post, "Tencent ramps up overseas AI push with wider roll-out of flagship Hy3 model," August 5, 2026. https://www.scmp.com/tech/tech-war/article/3362991/tencent-ramps-overseas-ai-push-wider-rollout-flagship-hy3-model
OpenRouter, "Tencent: Hy3," model listing, listed July 6, 2026. https://openrouter.ai/tencent/hy3
Tencent (@TencentGlobal), X post announcing global availability of Hy3, August 5, 2026. https://x.com/TencentGlobal/status/2085029369047626129

Editing: Tencent Hunyuan Hy3