Up to date
deepseek-ai DeepSeek License (deepseek-license) out 2 years ago
15.7B parameters (6 of 64 experts per token), text in, text out, 160K tokens of context.
Verified 2026-10-25 Rank #163 in LLMs · score 91.7 method v1 Sources 3 Downloads 30d 393.6k · #184 of 619 tracked
Picked by rule from 8 tracked pieces: the lab's own post, the most shared, a field report, the latest.
The 2 picks above, plus these 6. 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: 68 quantized builds and 84 Spaces.
Top Spaces running it: KBaba7/Quant ♥ 51 · FallnAI/Quantize-HF-Models ♥ 19 · bhaskartripathi/LLM_Quantization ♥ 16 · openfree/LLM_Quantization ♥ 16
Weights only, an estimate: leave room for the KV cache. Method
About 9.4 GB at INT4, 19 GB at FP8, 38 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 (2024-06-14), amber triangles are 8 dated articles (each links to it), grey ticks are repo commits.
93 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.
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: 15,706,484,224 |
| Size band | L |
| MoE | yes |
| Experts | 64 |
| Experts per token | 6 |
| Context length | 163,840 (config.json max_position_embeddings) |
| Papers | arXiv:2401.06066 |
| License | DeepSeek License (deepseek-license) (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-Coder-V2-Lite-Instruct: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 38 GB · Single high-end GPU (<80 GB) |
|---|---|
| FP8 (1 byte/param) | 19 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 9.4 GB · Single GPU (<24 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: 68finetune: 18adapter: 7
| Date | Commit | SHA |
|---|---|---|
| 2024-07-03 | Update README.md | e434a23 |
| 2024-06-24 | Update modeling_deepseek.py | 45d0aa4 |
| 2024-06-19 | Update README.md | ec228ab |
| 2024-06-18 | Update README.md | 21f7d9f |
| 2024-06-17 | Update README.md | b6ae3a9 |
| 2024-06-17 | Update README.md | 7389984 |
| 2024-06-16 | Create README.md | c48c7cf |
| 2024-06-14 | Remove unused file | 53268db |
| 2024-06-14 | Upload folder using huggingface_hub | 34397f9 |
| 2024-06-14 | initial commit | 29cd052 |
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 21:35 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:02 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.
