Up to date
deepseek-ai DeepSeek License (deepseek) out 3 years ago
236B parameters (6 of 160 experts per token), text in, text out, 160K tokens of context.
Verified 2026-10-25 Rank #689 in LLMs · score 72.3 method v1 Sources 3 Downloads 30d 12.7k · #488 of 619 tracked
Picked by rule from 8 tracked pieces: the lab's own post, the most shared, a field report, the latest.
The 4 picks above, plus these 4. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
Hugging Face does not say who downloads a model or with what. These are the proxies: API providers listing it, quantized builds, fine-tunes/adapters/merges, and Spaces that load it.
No API provider serves it, so its downloads are people running it themselves: 1 quantized builds and 15 Spaces.
Top Spaces running it: Justinrune/LLaMA-Factory ♥ 12 · artificialguybr/CanIRunThisLLM ♥ 9 · Csplk/moondream2-batch-processing ♥ 8 · concedo/WebTokenizer ♥ 5
Weights only, an estimate: leave room for the KV cache. Method
About 141 GB at INT4, 283 GB at FP8, 566 GB at BF16: weights only, so leave room for the KV cache. Every expert must be resident, so total parameters count.
Estimate params × bytes per param × 1.2 · how we estimate
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2024-04-22), amber triangles are 8 dated articles (each links to it), grey ticks are repo commits.
2 derivatives on the Hub. Which build to run, by runtime: the most downloaded in each class.
From the Hugging Face model tree; quantized builds in teal, fine-tunes, merges and adapters in greys.
Everything else, folded: the full tables, every source.
| Benchmark | Value | Setting / notes | Reported by | Date |
|---|---|---|---|---|
| author-reported BBH · Reason through hard problems | 78.9 | Base Model | model author | 2026-10-07 |
| author-reported BBH · Reason through hard problems | 81.3 | DeepSeek-V2 Chat (SFT) | model author | 2026-10-07 |
| author-reported BBH · Reason through hard problems | 79.7 | DeepSeek-V2 Chat (RL) | model author | 2026-10-07 |
| author-reported C-Eval · Work across languages | 81.7 | Base Model | model author | 2026-10-07 |
| author-reported C-Eval · Work across languages | 80.9 | DeepSeek-V2 Chat (SFT) | model author | 2026-10-07 |
| author-reported C-Eval · Work across languages | 78 | DeepSeek-V2 Chat (RL) | model author | 2026-10-07 |
| author-reported CMMLU · Work across languages | 84 | Base Model | model author | 2026-10-07 |
| author-reported CMMLU · Work across languages | 82.4 | DeepSeek-V2 Chat (SFT) | model author | 2026-10-07 |
| author-reported CMMLU · Work across languages | 81.6 | DeepSeek-V2 Chat (RL) | model author | 2026-10-07 |
| author-reported GSM8K · Reason through hard problems | 79.2 | Base Model | model author | 2026-10-07 |
| author-reported HumanEval · Write code with agents | 48.8 | Base Model | model author | 2026-10-07 |
| author-reported HumanEval · Write code with agents | 76.8 | DeepSeek-V2 Chat (SFT) | model author | 2026-10-07 |
| author-reported HumanEval · Write code with agents | 81.1 | DeepSeek-V2 Chat (RL) | model author | 2026-10-07 |
| author-reported LiveCodeBench · Write code with agents | 28.7 | DeepSeek-V2 Chat (SFT) | model author | 2026-10-07 |
| author-reported MATH-500 · Reason through hard problems | 43.6 | Base Model | model author | 2026-10-07 |
| author-reported MBPP · Write code with agents | 66.6 | Base Model | model author | 2026-10-07 |
| author-reported MBPP · Write code with agents | 70.4 | DeepSeek-V2 Chat (SFT) | model author | 2026-10-07 |
| author-reported MBPP · Write code with agents | 72 | DeepSeek-V2 Chat (RL) | model author | 2026-10-07 |
| author-reported MMLU · Reason through hard problems | 78.5 | Base Model | model author | 2026-10-07 |
| author-reported MMLU · Reason through hard problems | 78.4 | DeepSeek-V2 Chat (SFT) | model author | 2026-10-07 |
| author-reported MMLU · Reason through hard problems | 77.8 | DeepSeek-V2 Chat (RL) | model author | 2026-10-07 |
Labels are evidence, not verdicts: measured a third-party leaderboard row for this model; author-reported a number the model author published in the card or eval files; structural a deterministic signal from the name, config or chat template, or a decision-model estimate. Each chip links to its source. All use cases.
| Pipeline tag | text-generation |
|---|---|
| Library | transformers |
| Architectures | DeepseekV2ForCausalLM |
| Model type | deepseek_v2 |
| Dtype | bfloat16 |
| Params by dtype | BF16: 235,741,434,880 |
| Size band | XXL |
| MoE | yes |
| Experts | 160 |
| Experts per token | 6 |
| Context length | 163,840 (config.json max_position_embeddings) |
| Papers | arXiv:2311.18743, arXiv:2405.04434 |
| License | DeepSeek License (deepseek) (other) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · DeepseekV2ForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · DeepseekV2ForCausalLM · 773cb938fe9484a8f7e7e46fb1b1fe96412c59ed
vLLM registry (7050ef881ff0835f6890a47db1859a24c65481d6) · SGLang registry (773cb938fe9484a8f7e7e46fb1b1fe96412c59ed) · Matched by architecture class name against each project's registry on main; 'supported' means the architecture is registered, not that this checkpoint was tested.
Weights-only memory for DeepSeek-V2: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 566 GB · Multi-GPU node (<640 GB) |
|---|---|
| FP8 (1 byte/param) | 283 GB · Multi-GPU node (<640 GB) |
| INT4 (0.5 bytes/param) | 141 GB · Multi-GPU node (<640 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra. MoE: all experts must be resident, so total parameters count.
quantized: 1finetune: 1
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| mradermacher/DeepSeek-V2-i1-GGUF | quantized | GGUF | — | — | other | mradermacher | 0 | 900 |
| mradermacher/DeepSeek-V2-GGUF | finetune | GGUF | — | — | other | mradermacher | 0 | 0 |
| Date | Commit | SHA |
|---|---|---|
| 2024-06-08 | docs: update README.md | 4461458 |
| 2024-06-07 | docs: update README.md | f9b5644 |
| 2024-05-27 | training: fix type mismatch when training (#6) | 94cf30b |
| 2024-05-09 | Update code & readme | e0828e3 |
| 2024-05-08 | Update README.md | 3d3dec2 |
| 2024-05-08 | update citation info | 1e51912 |
| 2024-05-07 | Update README.md | 3b7037d |
| 2024-05-07 | Update README.md | ca159e6 |
| 2024-05-07 | Update README.md | d734100 |
| 2024-05-06 | Update model names | 44f7caf |
| 2024-05-06 | Update README.md | 3d445e7 |
| 2024-05-06 | Update README.md | 70cd796 |
| 2024-05-06 | Update README.md | e8c35ff |
| 2024-05-06 | Update README.md | 1142605 |
| 2024-05-06 | Update README.md | dd5954a |
| 2024-05-06 | Update README.md | b3333a6 |
| 2024-05-06 | Upload LICENSE | e63dbae |
| 2024-05-06 | Upload folder using huggingface_hub | 4c2b2d1 |
| 2024-04-22 | initial commit | b129598 |
Ranked by source signals (HN, Lobsters, authority domains, official posts) and recency. Summaries are generated from a stored copy of the article text; every sentence cites a span of that text, and nothing is shown that failed the citation check.
arXiv links found in the model card and its tags. "Peer-reviewed" is shown only with evidence: a published venue in OpenAlex (measured) or the authors' own journal reference on arXiv (reported).
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:41 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:13 UTC · Last verified 2026-10-25
Email alerts are not open yet; the feeds above are the same data. See feeds for the item schema.
