Up to date
deepseek-ai MIT out 21 months ago
32.8B parameters, text in, text out, 128K tokens of context. 1 provider serves it.
Verified 2026-10-25 Rank #84 in LLMs · score 102.1 method v1 Sources 3 Downloads 30d 350.4k · #198 of 619 tracked
Picked by rule from 35 tracked pieces: the lab's own post, the most shared, a field report, the latest.
Zyphra released ZAYA1-8B, a model that matches DeepSeek-R1 on math benchmarks while using less than 1 billion active parameters.
The 2 picks above, plus these 23. Top is the radar's score: discussion, source authority and the lab's own posts, fading with age.
Live hosted endpoints, cheapest first, USD per million tokens. Ties break on output price, then provider name.
| # | Cheapest live row | In /1M | Out /1M | Context | Quant | Seen |
|---|---|---|---|---|---|---|
| 1 | Featherless | — | — | — | — | Oct 25 |
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.
Served by 1 API provider; heavily run locally (148 quantized builds); a big base for builders (352 fine-tunes, adapters and merges).
Top Spaces running it: Vokturz/can-it-run-llm ♥ 1,053 · pliny-the-prompter/obliteratus ♥ 621 · llamameta/DeepSeek-R1-Chat-Assistant-Web-Search ♥ 260 · hadadrjt/ai ♥ 86
Weights only, an estimate: leave room for the KV cache. Method
About 20 GB at INT4, 39 GB at FP8, 79 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 (2025-01-20), amber triangles are 10 dated articles (each links to it), grey ticks are repo commits.
500 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 AIME 2024 · Reason through hard problems | 72.6 pass@1 | DeepSeek-R1-Distill-Qwen-32B | model author | 2026-10-07 |
| author-reported AIME 2024 · Reason through hard problems | 83.3 other | DeepSeek-R1-Distill-Qwen-32B | model author | 2026-10-07 |
| author-reported Codeforces · Write code with agents | 1691 rating | DeepSeek-R1-Distill-Qwen-32B | model author | 2026-10-07 |
| author-reported GPQA Diamond · Reason through hard problems | 62.1 pass@1 | DeepSeek-R1-Distill-Qwen-32B | model author | 2026-10-07 |
| author-reported LiveCodeBench · Write code with agents | 57.2 pass@1 | DeepSeek-R1-Distill-Qwen-32B | model author | 2026-10-07 |
| author-reported MATH-500 · Reason through hard problems | 94.3 pass@1 | DeepSeek-R1-Distill-Qwen-32B | 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.
Listed in LLMs
| Pipeline tag | text-generation |
|---|---|
| Library | transformers |
| Architectures | Qwen2ForCausalLM |
| Model type | qwen2 |
| Dtype | bfloat16 |
| Params by dtype | BF16: 32,763,876,352 |
| Size band | L |
| MoE | no |
| Context length | 131,072 (config.json max_position_embeddings) |
| Papers | arXiv:2501.12948 |
| License | MIT (mit) |
| Hub snapshot | 2026-10-25 |
vLLM: supported · Qwen2ForCausalLM · 7050ef881ff0835f6890a47db1859a24c65481d6SGLang: supported · Qwen2ForCausalLM · 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-R1-Distill-Qwen-32B: params × bytes/param × 1.2 overhead. KV cache and activations are extra. How this is computed.
| BF16 (2 bytes/param) | 79 GB · Single high-end GPU (<80 GB) |
|---|---|
| FP8 (1 byte/param) | 39 GB · Single high-end GPU (<80 GB) |
| INT4 (0.5 bytes/param) | 20 GB · Single GPU (<24 GB) |
Weights only (params × bytes/param × 1.2 overhead); KV cache and activations extra.
quantized: 148finetune: 91merge: 35adapter: 226
| Variant | Relation | Quant | Params | Band | License | Author | Likes | Downloads |
|---|---|---|---|---|---|---|---|---|
| bartowski/DeepSeek-R1-Distill-Qwen-32B-GGUF | quantized | GGUF | — | — | — | bartowski | 332 | 3.8M |
| unsloth/DeepSeek-R1-Distill-Qwen-32B-bnb-4bit | quantized | -bnb | 33.7B | L | apache-2.0 | unsloth | 29 | 1.3M |
| unsloth/DeepSeek-R1-Distill-Qwen-32B-GGUF | quantized | GGUF | — | — | apache-2.0 | unsloth | 171 | 1.3M |
| mlx-community/DeepSeek-R1-Distill-Qwen-32B-4bit | quantized | -4bit | 32.8B | L | — | mlx-community | 50 | 779.2k |
| mlx-community/DeepSeek-R1-Distill-Qwen-32B-MLX-8Bit | quantized | MLX | 32.8B | L | — | mlx-community | 18 | 612.5k |
| lmstudio-community/DeepSeek-R1-Distill-Qwen-32B-GGUF | quantized | GGUF | — | — | — | lmstudio-community | 40 | 594.8k |
| Valdemardi/DeepSeek-R1-Distill-Qwen-32B-AWQ | quantized | AWQ | 32.8B | L | apache-2.0 | Valdemardi | 38 | 492.1k |
| huihui-ai/DeepSeek-R1-Distill-Qwen-32B-abliterated | finetune | — | 32.8B | L | — | huihui-ai | 258 | 257.8k |
| JamAndTeaStudios/DeepSeek-R1-Distill-Qwen-32B-FP8-Dynamic | quantized | -FP8 | 32.8B | L | mit | JamAndTeaStudios | 2 | 263.3k |
| inarikami/DeepSeek-R1-Distill-Qwen-32B-AWQ | quantized | AWQ | 32.8B | L | mit | inarikami | 12 | 260.6k |
| RedHatAI/DeepSeek-R1-Distill-Qwen-32B-FP8-dynamic | quantized | -FP8 | 32.8B | L | mit | RedHatAI | 10 | 143.7k |
| mlx-community/DeepSeek-R1-Distill-Qwen-32B-bf16 | finetune | -bf16 | 32.8B | L | — | mlx-community | 4 | 109.4k |
| RedHatAI/DeepSeek-R1-Distill-Qwen-32B-quantized.w4a16 | quantized | -quant | 32.8B | L | mit | RedHatAI | 6 | 79.6k |
| unsloth/DeepSeek-R1-Distill-Qwen-32B | finetune | — | 32.8B | L | apache-2.0 | unsloth | 11 | 75.6k |
| cyberagent/DeepSeek-R1-Distill-Qwen-32B-Japanese | finetune | — | 32.8B | L | mit | cyberagent | 254 | 39.7k |
| qihoo360/TinyR1-32B-Preview | finetune | — | 32.8B | L | apache-2.0 | qihoo360 | 328 | 29.5k |
| Donnyed/DeepSeek-R1-Distill-Qwen-32B-Q4_K_M-GGUF | quantized | GGUF | — | — | — | Donnyed | 16 | 40k |
| mradermacher/DeepSeek-R1-Distill-Qwen-32B-GGUF | quantized | GGUF | — | — | mit | mradermacher | 2 | 37.5k |
| RedHatAI/DeepSeek-R1-Distill-Qwen-32B-quantized.w8a8 | quantized | -quant | 32.8B | L | mit | RedHatAI | 14 | 35.1k |
| mlx-community/DeepSeek-R1-Distill-Qwen-32B-6bit | quantized | — | 32.8B | L | — | mlx-community | 3 | 33.3k |
| mlx-community/DeepSeek-R1-Distill-Qwen-32B-3bit | quantized | — | 32.8B | L | — | mlx-community | 4 | 30k |
| Tongyi-Zhiwen/QwenLong-L1-32B | finetune | — | 32.8B | L | apache-2.0 | Tongyi-Zhiwen | 168 | 16.7k |
| unsloth/DeepSeek-R1-Distill-Qwen-32B-unsloth-bnb-4bit | quantized | -bnb | 32.8B | L | apache-2.0 | unsloth | 12 | 23.2k |
| Mungert/QwenLong-L1-32B-GGUF | quantized | GGUF | — | — | apache-2.0 | Mungert | 9 | 21.2k |
| DevQuasar-5/deepseek-ai.DeepSeek-R1-Distill-Qwen-32B-GGUF | quantized | GGUF | — | — | — | DevQuasar-5 | 2 | 20.7k |
| mradermacher/DeepSeek-R1-Distill-Qwen-32B-i1-GGUF | quantized | GGUF | — | — | mit | mradermacher | 6 | 18.3k |
| ggml-org/DeepSeek-R1-Distill-Qwen-32B-Q8_0-GGUF | quantized | Q8_0 | — | — | — | ggml-org | 7 | 17.8k |
| rinna/deepseek-r1-distill-qwen2.5-bakeneko-32b | merge | — | 32.8B | L | apache-2.0 | rinna | 7 | 15.9k |
| empirischtech/DeepSeek-R1-Distill-Qwen-32B-gptq-4bit | quantized | gptq | 32.8B | L | cc-by-4.0 | empirischtech | 16 | 14.6k |
| Skywork/Skywork-OR1-32B | finetune | — | 32.8B | L | — | Skywork | 22 | 13.7k |
| avoroshilov/DeepSeek-R1-Distill-Qwen-32B-GPTQ_4bit-128g | quantized | GPTQ | 32.8B | L | mit | avoroshilov | 0 | 14.6k |
| roleplaiapp/DeepSeek-R1-Distill-Qwen-32B-Q4_0-GGUF | quantized | Q4_0 | — | — | — | roleplaiapp | 6 | 12.7k |
| mlx-community/DeepSeek-R1-Distill-Qwen-32B-MLX-4Bit | quantized | MLX | 32.8B | L | — | mlx-community | 5 | 11.7k |
| nm-testing/DeepSeek-R1-Distill-Qwen-32B-NVFP4 | quantized | NVFP4 | 19.1B | L | mit | nm-testing | 3 | 11.7k |
| manarj789/DeepSeek-R1-Distill-Qwen-32B-Q8_0-GGUF | quantized | Q8_0 | — | — | — | manarj789 | 1 | 10.6k |
| mmnga/DeepSeek-R1-Distill-Qwen-32B-gguf | quantized | gguf | — | — | mit | mmnga | 3 | 9.5k |
| PKU-DS-LAB/FairyR1-32B | finetune | — | 32.8B | L | apache-2.0 | PKU-DS-LAB | 102 | 4k |
| roleplaiapp/DeepSeek-R1-Distill-Qwen-32B-Q2_K-GGUF | quantized | Q2_K | — | — | — | roleplaiapp | 2 | 8.8k |
| second-state/DeepSeek-R1-Distill-Qwen-32B-GGUF | quantized | GGUF | — | — | — | second-state | 0 | 8.6k |
| qihoo360/Light-R1-32B-DS | finetune | — | 32.8B | L | apache-2.0 | qihoo360 | 15 | 7.4k |
| Date | Commit | SHA |
|---|---|---|
| 2025-02-24 | Update bib info | 711ad2e |
| 2025-02-09 | Update tokenizer_config.json | 3865e12 |
| 2025-02-01 | Update README.md | b950d47 |
| 2025-01-26 | Update README.md | d66bcfc |
| 2025-01-23 | Update README.md | 10d6a03 |
| 2025-01-22 | Update tokenizer_config.json | ca24ee4 |
| 2025-01-22 | Update tokenizer_config.json | d5e5623 |
| 2025-01-21 | Add generation_config.json | 2d78713 |
| 2025-01-21 | add library tag for better code snippets and tags (#3) | 4569fd7 |
| 2025-01-20 | Release DeepSeek-R1 | 2a29ab1 |
| 2025-01-20 | Add files using upload-large-folder tool | 22b62e0 |
| 2025-01-20 | initial commit | f8e1f98 |
Hosted endpoints observed for DeepSeek-R1-Distill-Qwen-32B, sorted by input price (USD per million tokens). Each row names its observer and date; prices change daily, so check the source before quoting.
| Provider | Model ref | Context | Quant | $/M in | $/M out | tok/s | TTFT | Tools | Status | Observed |
|---|---|---|---|---|---|---|---|---|---|---|
| Featherless hf | deepseek-ai/DeepSeek-R1-Distill-Qwen-32B | — | — | — | — | — | — | — | live | 2026-10-25 |
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 21:33 UTC · Providers: 2026-10-24 21:44 UTC · Articles: 2026-10-24 21:45 UTC · Summary: 2026-10-07 15:00 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.
