# Ling-3.1-flash

> Source: https://aiwiki.ai/wiki/ling_3_1_flash
> Updated: 2026-10-15
> Categories: AI Models, Large Language Models, Open Source AI, Reasoning Models
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) - attribute to "AI Wiki (aiwiki.ai)"
> Cite as: AI Wiki. "Ling-3.1-flash." aiwiki.ai, 15 Oct 2026. https://aiwiki.ai/wiki/ling_3_1_flash
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution.

**Ling-3.1-flash** is a text [large language model](https://aiwiki.ai/wiki/large_language_model) from [InclusionAI](https://aiwiki.ai/wiki/inclusionai), part of Ant Group's Ling family. The publisher describes it as a hybrid reasoning model with approximately 560 billion total parameters and 25 billion activated per token, intended for software development, office work and tasks involving repeated tool use.[2]

The model has downloadable weights under an [MIT license](https://aiwiki.ai/wiki/mit_license).[1] Its advertised maximum context is 1 million tokens, but that figure differs from the 262,144-position setting in the released configuration and the 256K limit documented for the official hosted service. These limits should be distinguished when selecting a deployment.[2][3]

## Architecture and specifications

Ling-3.1-flash combines linear attention with [mixture-of-experts](https://aiwiki.ai/wiki/mixture_of_experts) feed-forward layers. Its backbone has 72 layers: seven Kimi Delta Attention (KDA) layers followed by one gated [Multi-head Latent Attention](https://aiwiki.ai/wiki/multi-head_latent_attention) layer in each eight-layer group. The initial four feed-forward layers are dense; later ones route tokens through experts.[4]

| Property | Released specification |
| --- | --- |
| Architecture implementation | `BailingMoeV3ForCausalLM` |
| Total parameters, publisher's count | Approximately 560B |
| Activated parameters per token, publisher's count | Approximately 25B |
| Backbone layers | 72 |
| Attention-layer composition | 63 KDA + 9 gated MLA |
| Initial dense feed-forward layers | 4 |
| Routed experts per MoE layer | 512 |
| Selected routed experts per token | 8 |
| Shared experts per MoE layer | 1 |
| Attention heads | 64 |
| Hidden size | 4096 |
| Expert intermediate size | 1280 |
| Dense intermediate size | 10240 |
| Vocabulary size | 157184 |
| Additional next-token prediction layers | 1 |
| Configuration's maximum positions | 262144 |
| Published checkpoint quantization | FP8 E4M3, 128 x 128 weight blocks |

The parameter estimates are the publisher's figures. The configuration supplies the remaining numerical specifications.[2][3]

The shared expert is added to the routed-expert output. The extra prediction layer is separate from the 72-layer backbone and supports [multi-token prediction](https://aiwiki.ai/wiki/multi_token_prediction). The reference implementation selects MLA attention for that extra layer.[4]

The checkpoint's [FP8](https://aiwiki.ai/wiki/fp8) configuration specifies dynamic activation quantization and excludes selected components, including embeddings, the output head and router-related projections, from conversion.[3]

## Context length

The publisher's documentation describes a model-level upper limit of 1M tokens while listing the official service at 256K. Its model-selection table also lists 256K for Ling-3.1-flash.[2]

For the downloadable checkpoint, `max_position_embeddings` is `262144` and `rope_scaling` is `null`. The SGLang cookbook likewise gives a 262,144-token context limit. The advertised 1M figure should therefore not be read as confirmation that the published default configuration or every hosted endpoint accepts a million-token request.[3][5]

## Reasoning and tool use

The released chat template supports reasoning-enabled and reasoning-disabled requests. It defaults to thinking on unless an explicit setting overrides it. The model card recommends `temperature=1.0`, `top_p=0.95` and `top_k=20` with thinking enabled.[1][6]

For example, a compatible chat-completions request can disable thinking with:

```json
{
  "chat_template_kwargs": {
    "enable_thinking": false
  }
}
```

The template serializes tool definitions and tool-call arguments into its own role and XML-like delimiters. Applications should use the checkpoint's template and a serving implementation that understands it rather than substitute another model's formatting.[6]

SGLang's `ling3` parser separates `reasoning_content` from final `content`. Its tool parser returns structured `tool_calls`; a follow-up request supplies each result with the matching `tool_call_id`.[5]

The publisher also describes LingCode, a coding harness based on Pi for long-running tasks and repeated context compaction. The model card says the team is considering releasing LingCode as open source; this statement is not a release of the harness.[1]

## Self-hosting

### SGLang

The [SGLang](https://aiwiki.ai/wiki/sglang) cookbook provides FP8 deployment recipes verified on eight H200 GPUs or four GB300 GPUs. The corresponding tensor-parallel and expert-parallel settings are `--tp 8 --ep 8` and `--tp 4 --ep 4`. Recipes for H100, H800, H20-3e and B200 are listed as unverified.[5]

The day-0 image is `lmsysorg/sglang:dev-ling-3.1-flash`. Its `NEXTN` option enables built-in [speculative decoding](https://aiwiki.ai/wiki/speculative_decoding). The cookbook warns that NEXTN sets a default maximum of 48 running requests unless the operator supplies another value. It also documents optional Hopper decode optimizations and Mooncake-backed HiCache prefix storage.[5]

These are particular tested configurations, not a universal minimum GPU count. The cookbook distinguishes verified recipes from experimental settings and notes that its prefill/decode-disaggregated deployment has not been verified.[5]

### vLLM

The model card includes a [vLLM](https://aiwiki.ai/wiki/vllm) launch recipe using tensor parallelism of eight, expert parallelism, the `ling3` tool and reasoning parsers, prefix caching and MTP speculative decoding. It also enables `--trust-remote-code`; operators should review the repository's custom implementation before enabling that option.[1]

## Evaluation

[Artificial Analysis](https://aiwiki.ai/wiki/artificial_analysis) listed a score of 41 on its Intelligence Index v4.3.2 for Ling 3.1 Flash when checked on 16 October 2026. That index combines ten evaluations, including agentic knowledge work, coding, long-context reasoning and knowledge reliability. The same page displayed output speed of 212.6 tokens per second and approximately 220 million output tokens used across the index evaluation.[7]

These are the evaluator's hosted-model measurements. They do not establish the throughput of a different local GPU configuration, and the index score is not a percentage accuracy across all possible tasks.[7]

The publisher reports evaluations covering GDPVal-AA V2.1, HealthBench Professional, FrontierSWE and Terminal-Bench 4.0. Those benchmark names describe different task sets and should not be treated as interchangeable measures of general reliability.[1]

## Training-content disclosure

The publisher's training-content summary identifies Ling-3.1-flash and describes a multilingual text corpus exceeding 10 trillion tokens. It names English and Chinese as major languages, reports coverage of other languages, and gives approximately July 2026 as the latest data-acquisition date. The corpus includes public material, licensed data and synthetic text.[8]

The summary identifies [Common Crawl](https://aiwiki.ai/wiki/common_crawl) among the large public datasets. It also describes academic papers, technical material, programming questions and answers, code, books and other web text. These are broad source categories, not a complete downloadable dataset inventory.[8]

The disclosure is a provider statement about training content, not an independent audit of model behavior. InclusionAI's documentation repository says the summaries neither distribute the underlying datasets nor grant rights to third-party content. It also cautions that publishing a summary is not a certification or endorsement by an authority.[9]

## References

1. InclusionAI. [Ling-3.1-flash model card](https://huggingface.co/inclusionAI/Ling-3.1-flash). Accessed 16 October 2026.
2. Ant Ling. [Ling model documentation](https://developer.ant-ling.com/en/docs/models/ling), including model-selection and context-length information. Accessed 16 October 2026.
3. InclusionAI. [Ling-3.1-flash configuration](https://huggingface.co/inclusionAI/Ling-3.1-flash/blob/main/config.json). Accessed 16 October 2026.
4. InclusionAI. [Ling-3.1-flash reference implementation](https://huggingface.co/inclusionAI/Ling-3.1-flash/blob/main/modeling_bailing_moe_v3.py). Accessed 16 October 2026.
5. SGLang. [Ling-3.1-flash deployment cookbook](https://docs.sglang.io/cookbook/autoregressive/InclusionAI/Ling-3.1-flash). Accessed 16 October 2026.
6. InclusionAI. [Ling-3.1-flash chat template](https://huggingface.co/inclusionAI/Ling-3.1-flash/blob/main/chat_template.jinja). Accessed 16 October 2026.
7. Artificial Analysis. [Ling 3.1 Flash intelligence and performance analysis](https://artificialanalysis.ai/models/ling-3-1-flash). Accessed 16 October 2026.
8. Ant Intelligence (Hang Zhou) Technology Co., Ltd. [Ling-3.1 training-content summary](https://huggingface.co/inclusionAI/AI-Transparency/resolve/main/Ling-3.1_TDS-Summary.pdf), sections 1.2-2.1. Accessed 16 October 2026.
9. InclusionAI. [AI-Transparency documentation index](https://huggingface.co/inclusionAI/AI-Transparency). Accessed 16 October 2026.