Up to date
deepseek-ai DeepSeek License (deepseek) 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 #418 in LLMs · score 79 method v1 Sources 3 Downloads 30d 70k · #355 of 619 tracked
Picked by rule from 14 tracked pieces: the lab's own post, the most shared, a field report, the latest.
The 2 picks above, plus these 12. 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: 42 quantized builds and 32 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 (2024-05-15), amber triangles are 7 dated articles (each links to it), grey ticks are repo commits.
55 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 | 44.1 | DeepSeek-V2-Lite (MoE-16B) Base | model author | 2026-10-07 |
| author-reported BBH · Reason through hard problems | 48.1 | DeepSeek-V2-Lite 16B Chat (SFT) | model author | 2026-10-07 |
| author-reported C-Eval · Work across languages | 60.3 | DeepSeek-V2-Lite (MoE-16B) Base | model author | 2026-10-07 |
| author-reported C-Eval · Work across languages | 60.1 | DeepSeek-V2-Lite 16B Chat (SFT) | model author | 2026-10-07 |
| author-reported CMMLU · Work across languages | 64.3 | DeepSeek-V2-Lite (MoE-16B) Base | model author | 2026-10-07 |
| author-reported CMMLU · Work across languages | 62.5 | DeepSeek-V2-Lite 16B Chat (SFT) | model author | 2026-10-07 |
| author-reported GSM8K · Reason through hard problems | 41.1 | DeepSeek-V2-Lite (MoE-16B) Base | model author | 2026-10-07 |
| author-reported GSM8K · Reason through hard problems | 72 | DeepSeek-V2-Lite 16B Chat (SFT) | model author | 2026-10-07 |
| author-reported HumanEval · Write code with agents | 29.9 | DeepSeek-V2-Lite (MoE-16B) Base | model author | 2026-10-07 |
| author-reported HumanEval · Write code with agents | 57.3 | DeepSeek-V2-Lite 16B Chat (SFT) | model author | 2026-10-07 |
| author-reported MATH-500 · Reason through hard problems | 17.1 | DeepSeek-V2-Lite (MoE-16B) Base | model author | 2026-10-07 |
| author-reported MATH-500 · Reason through hard problems | 27.9 | DeepSeek-V2-Lite 16B Chat (SFT) | model author | 2026-10-07 |
| author-reported MBPP · Write code with agents | 43.2 | DeepSeek-V2-Lite (MoE-16B) Base | model author | 2026-10-07 |
| author-reported MBPP · Write code with agents | 45.8 | DeepSeek-V2-Lite 16B Chat (SFT) | model author | 2026-10-07 |
| author-reported MMLU · Reason through hard problems | 58.3 | DeepSeek-V2-Lite (MoE-16B) Base | model author | 2026-10-07 |
| author-reported MMLU · Reason through hard problems | 55.7 | DeepSeek-V2-Lite 16B Chat (SFT) | 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: 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: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-Lite-Chat: 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: 42adapter: 5finetune: 8
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| mradermacher/DeepSeek-V2-Lite-Chat-GGUF | quantized | GGUF | — | — | other | mradermacher | 7 | 64k |
| legraphista/DeepSeek-V2-Lite-Chat-IMat-GGUF | quantized | GGUF | — | — | — | legraphista | 12 | 63.7k |
| second-state/DeepSeek-V2-Lite-Chat-GGUF | quantized | GGUF | — | — | other | second-state | 6 | 44.2k |
| mradermacher/DeepSeek-V2-Lite-Chat-i1-GGUF | quantized | GGUF | — | — | other | mradermacher | 1 | 38.4k |
| gaianet/DeepSeek-V2-Lite-Chat-GGUF | quantized | GGUF | — | — | other | gaianet | 0 | 10.3k |
| tensorblock/DeepSeek-V2-Lite-Chat-GGUF | quantized | GGUF | — | — | other | tensorblock | 0 | 5.2k |
| ggfox00000/deepseek-v2-lite-chat-gguf | quantized | gguf | — | — | other | ggfox00000 | 0 | 4k |
| phi0112358/DeepSeek-V2-Lite-Chat-Q5_K_M-GGUF | quantized | GGUF | — | — | other | phi0112358 | 0 | 3.9k |
| itlwas/DeepSeek-V2-Lite-Chat-Q4_K_M-GGUF | quantized | GGUF | — | — | other | itlwas | 0 | 2.5k |
| malaiwah/DeepSeek-V2-Lite-Chat-SIQ-K4K3 | quantized | — | 4.36B | M | other | malaiwah | 0 | 1.7k |
| zhentaoyu/DeepSeek-V2-Lite-Chat-Q4_0-GGUF | quantized | Q4_0 | — | — | other | zhentaoyu | 1 | 1.5k |
| YorkieOH10/DeepSeek-V2-Lite-Chat-Q4_K_M-GGUF | quantized | GGUF | — | — | other | YorkieOH10 | 0 | 1.4k |
| Zhu149248/DeepSeek-V2-Lite-Chat-ReMoE-GGUF | quantized | GGUF | — | — | other | Zhu149248 | 1 | 1.3k |
| Aldaris/DeepSeek-V2-Lite-Chat-Q4_K_M-GGUF | quantized | GGUF | — | — | other | Aldaris | 0 | 1.1k |
| pipilok/DeepSeek-V2-Lite-Chat-Q4_0_4_8-GGUF | quantized | GGUF | — | — | other | pipilok | 0 | 1k |
| phi0112358/DeepSeek-V2-Lite-Chat-Q4_0-GGUF | quantized | Q4_0 | — | — | other | phi0112358 | 0 | 905 |
| deevade/DeepSeek-V2-Lite-Chat-finetuned | adapter | GGUF | — | — | other | deevade | 1 | 703 |
| matrixportalx/DeepSeek-V2-Lite-Chat-Q4_0-GGUF | quantized | Q4_0 | — | — | other | matrixportalx | 2 | 596 |
| Jianping746/DeepSeek-V2-Lite-Chat-Q4_K_M-GGUF | quantized | GGUF | — | — | other | Jianping746 | 0 | 664 |
| slowfastai/DeepSeek-V2-Lite-Chat-bnb-4bit | quantized | -bnb | 16.2B | L | — | slowfastai | 0 | 624 |
| wuchen01/DeepSeek-V2-Lite-Chat-All-LoRA | adapter | — | — | — | other | wuchen01 | 0 | 611 |
| nightcatsong/DeepSeek-V2-Lite-Chat-Q8_0-GGUF | quantized | Q8_0 | — | — | other | nightcatsong | 0 | 540 |
| ijohn07/DeepSeek-V2-Lite-Chat-Q8_0-GGUF | quantized | Q8_0 | — | — | other | ijohn07 | 0 | 513 |
| Jianping746/DeepSeek-V2-Lite-Chat-Q6_K-GGUF | quantized | Q6_K | — | — | other | Jianping746 | 0 | 503 |
| alexander583/DeepSeek-V2-Lite-Chat-Q4-mlx | quantized | mlx | 15.7B | L | other | alexander583 | 0 | 487 |
| phi0112358/DeepSeek-V2-Lite-Chat-Q4_K_M-GGUF | quantized | GGUF | — | — | other | phi0112358 | 0 | 477 |
| ijohn07/DeepSeek-V2-Lite-Chat-Q6_K-GGUF | quantized | Q6_K | — | — | other | ijohn07 | 0 | 380 |
| NexesQuants/DeepSeek-V2-Lite-Chat-Q8_0-GGUF | quantized | Q8_0 | — | — | other | NexesQuants | 0 | 370 |
| Chang-chih/leviathan-16B-final | quantized | GGUF | — | — | apache-2.0 | Chang-chih | 1 | 284 |
| Lansechen/deepseek-v2-lite-16b-chat-R1-Distill-batch8-numinamath | finetune | — | 15.7B | L | — | Lansechen | 1 | 280 |
| YorkieOH10/DeepSeek-V2-Lite-Chat-Q8_0-GGUF | quantized | Q8_0 | — | — | other | YorkieOH10 | 0 | 317 |
| JenniSD/DeepSeek-V2-Lite-Chat-Q4_K_M-GGUF | quantized | GGUF | — | — | other | JenniSD | 0 | 315 |
| Lucy-in-the-Sky/DeepSeek-V2-Lite-Chat-Q4_K_M-GGUF | quantized | GGUF | — | — | other | Lucy-in-the-Sky | 0 | 305 |
| third0x/DeepSeek-V2-Lite-Chat-Q4_K_M-GGUF | quantized | GGUF | — | — | other | third0x | 0 | 304 |
| phi0112358/DeepSeek-V2-Lite-Chat-Q8_0-GGUF | quantized | Q8_0 | — | — | other | phi0112358 | 0 | 302 |
| Lansechen/deepseek-v2-lite-16b-chat-R1-Distill-bs17k-batch32 | finetune | — | 15.7B | L | — | Lansechen | 1 | 241 |
| RangerX/DeepSeek-V2-Lite-Chat-REAP-Pruned-ratio-0.3 | finetune | — | 11.4B | L | — | RangerX | 1 | 227 |
| Lansechen/deepseek-v2-lite-16b-chat-R1-Distill-batch16-lora-numinamath | finetune | — | — | — | — | Lansechen | 1 | 223 |
| third0x/DeepSeek-V2-Lite-Chat-Q8_0-GGUF | quantized | Q8_0 | — | — | other | third0x | 0 | 250 |
| Lucy-in-the-Sky/DeepSeek-V2-Lite-Chat-Q2_K-GGUF | quantized | Q2_K | — | — | other | Lucy-in-the-Sky | 0 | 238 |
| Date | Commit | SHA |
|---|---|---|
| 2024-06-25 | Update modeling_deepseek.py | 8586474 |
| 2024-06-08 | docs: update README.md | ef87e36 |
| 2024-06-07 | docs: update README.md | 003ee92 |
| 2024-05-17 | Update README.md | be9443d |
| 2024-05-17 | Upload LICENSE | d52a9ad |
| 2024-05-17 | Update README.md | 8edea68 |
| 2024-05-17 | Update README.md | 46310cf |
| 2024-05-17 | Update README.md | e91ebd8 |
| 2024-05-17 | Update README.md | 4621eeb |
| 2024-05-17 | Update README.md | 89a2dbc |
| 2024-05-16 | README.md | 1e90cb3 |
| 2024-05-16 | Update README.md | 935c70a |
| 2024-05-16 | Update README.md | 1d9f2af |
| 2024-05-16 | Update README.md | 7c56d86 |
| 2024-05-16 | Update README.md | 08529b4 |
| 2024-05-15 | Upload folder using huggingface_hub | a6f88f2 |
| 2024-05-15 | Update README.md | 198c4d1 |
| 2024-05-15 | Create README.md | ed322a8 |
| 2024-05-15 | initial commit | 148cc0a |
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 21:38 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:09 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.
