Up to date
deepseek-ai DeepSeek License (deepseek-license) out 3 years ago
6.74B parameters, text in, text out, 16K tokens of context.
Verified 2026-10-25 Rank #725 in LLMs · score 71.7 method v1 Sources 3 Downloads 30d 113.2k · #306 of 619 tracked
Picked by rule from 9 tracked pieces: the lab's own post, the most shared, a field report, the latest.
The 2 picks above, plus these 7. 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: 33 quantized builds and 54 Spaces.
Top Spaces running it: KBaba7/Quant ♥ 51 · Yeyito/llm_contamination_detector ♥ 27 · bartar/tokenizers ♥ 19 · FallnAI/Quantize-HF-Models ♥ 19
Weights only, an estimate: leave room for the KV cache. Method
About 4.0 GB at INT4, 8.1 GB at FP8, 16 GB at BF16: weights only, so leave room for the KV cache.
Estimate params × bytes per param × 1.2 · how we estimate
Weights, every dated article, and repo commits on one rail.
Teal is the weights (2023-10-23), amber triangles are 8 dated articles (each links to it), grey ticks are repo commits.
101 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.
The model was trained from scratch on 2T tokens to support project-level code completion and infilling.[1]
Generated 2026-10-07 15:16 UTC by parasail-gemma-4-26b-a4b-it; each numeral points at the fact it rests on. Not a verdict.
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.
Listed in LLMs
| Pipeline tag | text-generation |
|---|---|
| Library | transformers |
| Architectures | LlamaForCausalLM |
| Model type | llama |
| Dtype | bfloat16 |
| Params by dtype | BF16: 6,740,512,768 |
| Size band | M |
| MoE | no |
| Context length | 16,384 (config.json max_position_embeddings) |
| License | DeepSeek License (deepseek-license) (other) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · LlamaForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · LlamaForCausalLM · 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-6.7b-base: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 16 GB · Single GPU (<24 GB) |
|---|---|
| FP8 (1 byte/param) | 8.1 GB · Single GPU (<24 GB) |
| INT4 (0.5 bytes/param) | 4.0 GB · Single GPU (<24 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
quantized: 33finetune: 10adapter: 53merge: 5
| Date | Commit | SHA |
|---|---|---|
| 2024-03-19 | Adding `safetensors` variant of this model (#3) | ce2207a |
| 2023-11-05 | Update README.md | 1bd0bc7 |
| 2023-11-05 | Update config.json | 1be74df |
| 2023-11-04 | Update README.md | 9f75359 |
| 2023-11-03 | Update README.md | 7ecf1fb |
| 2023-11-02 | Update README.md | 9b8f9d7 |
| 2023-11-01 | Create README.md | 4da5b5b |
| 2023-10-28 | upload models | c9b068f |
| 2023-10-28 | Create LICENSE | 898c01a |
| 2023-10-23 | initial commit | 63f393d |
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.
| Title | Source | Published | First seen | Signals |
|---|---|---|---|---|
| DeepSeek v4.1 Flash | twitter.com | 2026-09-10 | 2026-10-07 | hn |
| DeepSeek V4 Flash 0731 Intelligence, Performance and Price Analysis | artificialanalysis.ai | 2026-07-31 | 2026-10-07 | hn |
| DeepSeek OCR | github.com | 2025-10-20 | 2026-10-07 | hn |
| The path to open-sourcing the DeepSeek inference engine | github.com | 2025-04-14 | 2026-10-07 | hn |
| DeepSeek Open Infra: Open-Sourcing 5 AI Repos in 5 Days | github.com | 2025-02-21 | 2026-10-07 | hn |
| OpenAI says it has evidence DeepSeek used its model to train competitor | ft.com | 2025-01-29 | 2026-10-07 | hn |
| DeepSeek releases Janus Pro, a text-to-image generator [pdf] | github.com | 2025-01-27 | 2026-10-07 | hn |
| DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via RL | arxiv.org | 2025-01-25 | 2026-10-07 | hn |
| DeepSeek-R1 | github.com | 2025-01-20 | 2026-10-07 | hn |
Hub snapshot: 2026-10-25 · HF API: 2026-10-24 21:43 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:46 UTC · Summary: 2026-10-07 15:16 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.
