- Size
- 685B parameterscounted from the safetensors header on the Hub
- Licence
- unknown, MIT LicenseOur reading, check the licence: You can use, change and redistribute the model for any purpose, including commercially. Keep the copyright notice and the licence text with copies. You do not have to share your changes.
- On the Hub since
- 28 May 2025the date the repo was created
- Context length
- 160k tokensfrom the model's config
- Kind
- Base model (no base model declared on the Hub)DeepseekV3ForCausalLM
- On Hugging Face
- deepseek-ai/DeepSeek-R1-0528143k downloads in the last 30 days, 2.5k likes
Scores
- default setting76.3%±2.429 May 2025, log
Epoch's name: DeepSeek-R1-0528. Best closed model on this set: gpt-6.2-astra_max, 96.8% (OpenAI). All 341 results
- default setting96.6%±0.429 May 2025, log
Epoch's name: DeepSeek-R1-0528. Best closed model on this set: gpt-5-2025-08-07_high, 98.1% (OpenAI). All 108 results
- default setting66.4%±4.929 May 2025, log
Epoch's name: DeepSeek-R1-0528. Best closed model on this set: claude-fable-5_high, 100.0% (Anthropic). All 318 results
- Edit format: diff71.4%6 Jun 2025, aider.chat
Epoch's name: DeepSeek-R1-0528. Best closed model on this set: gpt-5-2025-08-07_high, 88.0% (OpenAI). All 72 results
- default setting21.2%arcprize.org
Epoch's name: DeepSeek-R1-0528. Best closed model on this set: claude-fable-5_max, 98.5% (Anthropic). All 318 results
- default setting1.1%via Epoch AI
Epoch's name: DeepSeek-R1-0528. Best closed model on this set: gpt-6-astra_max, 95.0% (OpenAI). All 299 results
- default setting35.1%DeepResearchBench Leaderboard
Epoch's name: DeepSeek-R1-0528. Best closed model on this set: claude-opus-4-6_high, 55.3% (Anthropic). All 41 results
- default setting75.0%Fiction.live leaderboard
Epoch's name: chutes/DeepSeek-R1-0528. Best closed model on this set: gpt-5-2025-08-07_medium, 97.2% (OpenAI). All 61 results
- default setting81.9%lechmazur/writing Github repository
Epoch's name: DeepSeek-R1-0528. Best closed model on this set: gpt-5-2025-08-07_medium, 86.0% (OpenAI). All 49 results
- default setting53.8%METR - Measuring AI Ability to Complete Long Tasks
Epoch's name: DeepSeek-R1-0528. Best closed model on this set: claude-mythos-preview-early, 85.2% (Anthropic). All 49 results
- default setting40.8%simple-bench.com
Epoch's name: DeepSeek-R1-0528. Best closed model on this set: claude-fable-5-2_unknown, 94.9% (Anthropic). All 125 results
- default setting41.6%±0.0htihle.github.io
Epoch's name: DeepSeek-R1-0528. Best closed model on this set: gpt-6-astra_promax, 93.6% (OpenAI). All 173 results
Claimed by the maker Claimed
The model card has no claimed scores in a machine-readable form (the card's model-index), so none are shown. The maker may give scores in the card text: read the card.
Other views, which we link to and do not copy: look this model up on Artificial Analysis and the Arena leaderboard.
Independent scores from Epoch AI (CC-BY 4.0), fetched 19 Nov 2026: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks’ [online resource]. Each row links its own source. What the badges mean.
Run it yourself
Memory needed = weights + context cache + 0.50 GB for the runtime. Context cache at 8k tokens: 0.54 GB, worked out from the model's config (deepseek-ai/DeepSeek-R1-0528), at 16-bit precision. Choose a machine above to see a verdict for each format. Method
GGUF for llama.cpp, Ollama, LM Studio and others
unsloth/DeepSeek-R1-0528-GGUF (converted by unsloth), 13k downloads. Sizes are the real files on the Hub, all parts added up.
| Format | Size | Memory needed | Get it |
|---|---|---|---|
| UD-IQ3_XXS | 254.2 GBfile size, 6 parts | 255.2 GB254.2 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:UD-IQ3_XXSDownload hf download unsloth/DeepSeek-R1-0528-GGUF --include "UD-IQ3_XXS/DeepSeek-R1-0528-UD-IQ3_XXS-*-of-00006.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| Q3_K_S | 269.8 GBfile size, 6 parts | 270.9 GB269.8 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:Q3_K_SDownload hf download unsloth/DeepSeek-R1-0528-GGUF --include "Q3_K_S/DeepSeek-R1-0528-Q3_K_S-*-of-00006.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| UD-Q3_K_XL | 275.6 GBfile size, 7 parts | 276.6 GB275.6 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:UD-Q3_K_XLDownload hf download unsloth/DeepSeek-R1-0528-GGUF --include "UD-Q3_K_XL/DeepSeek-R1-0528-UD-Q3_K_XL-*-of-00007.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| Q3_K_M | 297.9 GBfile size, 7 parts | 298.9 GB297.9 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:Q3_K_MDownload hf download unsloth/DeepSeek-R1-0528-GGUF --include "Q3_K_M/DeepSeek-R1-0528-Q3_K_M-*-of-00007.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| IQ4_XS | 333.1 GBfile size, 8 parts | 334.2 GB333.1 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:IQ4_XSDownload hf download unsloth/DeepSeek-R1-0528-GGUF --include "IQ4_XS/DeepSeek-R1-0528-IQ4_XS-*-of-00008.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| IQ4_NL | 352.6 GBfile size, 8 parts | 353.6 GB352.6 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:IQ4_NLDownload hf download unsloth/DeepSeek-R1-0528-GGUF --include "IQ4_NL/DeepSeek-R1-0528-IQ4_NL-*-of-00008.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| Q4_0 | 353.5 GBfile size, 8 parts | 354.5 GB353.5 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:Q4_0Download hf download unsloth/DeepSeek-R1-0528-GGUF --include "Q4_0/DeepSeek-R1-0528-Q4_0-*-of-00008.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| Q4_K_S | 354.4 GBfile size, 8 parts | 355.4 GB354.4 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:Q4_K_SDownload hf download unsloth/DeepSeek-R1-0528-GGUF --include "Q4_K_S/DeepSeek-R1-0528-Q4_K_S-*-of-00008.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| UD-Q4_K_XL | 357.6 GBfile size, 8 parts | 358.7 GB357.6 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:UD-Q4_K_XLDownload hf download unsloth/DeepSeek-R1-0528-GGUF --include "UD-Q4_K_XL/DeepSeek-R1-0528-UD-Q4_K_XL-*-of-00008.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| Q4_K_M | 377.1 GBfile size, 9 parts | 378.2 GB377.1 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:Q4_K_MDownload hf download unsloth/DeepSeek-R1-0528-GGUF --include "Q4_K_M/DeepSeek-R1-0528-Q4_K_M-*-of-00009.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| Q4_1 | 391.5 GBfile size, 9 parts | 392.6 GB391.5 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:Q4_1Download hf download unsloth/DeepSeek-R1-0528-GGUF --include "Q4_1/DeepSeek-R1-0528-Q4_1-*-of-00009.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| Q5_K_S | 430.5 GBfile size, 10 parts | 431.5 GB430.5 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:Q5_K_SDownload hf download unsloth/DeepSeek-R1-0528-GGUF --include "Q5_K_S/DeepSeek-R1-0528-Q5_K_S-*-of-00010.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| Q5_K_M | 443.1 GBfile size, 10 parts | 444.1 GB443.1 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:Q5_K_MDownload hf download unsloth/DeepSeek-R1-0528-GGUF --include "Q5_K_M/DeepSeek-R1-0528-Q5_K_M-*-of-00010.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| UD-Q5_K_XL | 448.3 GBfile size, 10 parts | 449.3 GB448.3 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:UD-Q5_K_XLDownload hf download unsloth/DeepSeek-R1-0528-GGUF --include "UD-Q5_K_XL/DeepSeek-R1-0528-UD-Q5_K_XL-*-of-00010.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| Q6_K | 513.2 GBfile size, 12 parts | 514.2 GB513.2 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:Q6_KDownload hf download unsloth/DeepSeek-R1-0528-GGUF --include "Q6_K/DeepSeek-R1-0528-Q6_K-*-of-00012.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| UD-Q6_K_XL | 535.5 GBfile size, 12 parts | 536.5 GB535.5 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:UD-Q6_K_XLDownload hf download unsloth/DeepSeek-R1-0528-GGUF --include "UD-Q6_K_XL/DeepSeek-R1-0528-UD-Q6_K_XL-*-of-00012.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| Q8_0 | 664.3 GBfile size, 15 parts | 665.3 GB664.3 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:Q8_0Download hf download unsloth/DeepSeek-R1-0528-GGUF --include "Q8_0/DeepSeek-R1-0528-Q8_0-*-of-00015.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| UD-Q8_K_XL | 727.0 GBfile size, 16 parts | 728.0 GB727.0 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:UD-Q8_K_XLDownload hf download unsloth/DeepSeek-R1-0528-GGUF --include "UD-Q8_K_XL/DeepSeek-R1-0528-UD-Q8_K_XL-*-of-00016.gguf" --local-dir DeepSeek-R1-0528-GGUF |
| BF16 | 1250.1 GBfile size, 30 parts | 1251.1 GB1250.1 + 0.54 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/DeepSeek-R1-0528-GGUF:BF16Download hf download unsloth/DeepSeek-R1-0528-GGUF --include "BF16/DeepSeek-R1-0528-BF16-*-of-00030.gguf" --local-dir DeepSeek-R1-0528-GGUF |
MLX for Apple silicon
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| MLX 3bitmlx-community/DeepSeek-R1-0528-3bit | 273.6 GBfile size | 274.6 GB273.6 + 0.54 + 0.50 | Download page |
| MLX 5bitmlx-community/DeepSeek-R1-0528-5bit | 429.8 GBfile size | 430.8 GB429.8 + 0.54 + 0.50 | Download page |
Original weights size from the safetensors header, for vLLM, SGLang, Transformers
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| F8_E4M3original weights | 641.3 GBfile size | 642.3 GB641.3 + 0.54 + 0.50 |
hf download deepseek-ai/DeepSeek-R1-0528 --local-dir DeepSeek-R1-0528Other conversions (15; sizes not read)
- amd/DeepSeek-R1-0528-MXFP4 Other, mxfp4, 3.1k downloads
- amd/DeepSeek-R1-0528-MXFP4-MTP-MoEFP4 Other, mxfp4, 1.8k downloads
- amd/DeepSeek-R1-0528-MXFP4-v2 Other, mxfp4, 591 downloads
- m8than/DeepSeek-R1-0528-E2M1 Other, 251 downloads
- unsloth/DeepSeek-R1-0528 FP8, 242 downloads
- RedHatAI/DeepSeek-R1-0528-quantized.w4a16 GPTQ, w4a16, 178 downloads
- QuantTrio/DeepSeek-R1-0528-GPTQ-Int4-Int8Mix-Medium GPTQ, int4, 163 downloads
- QuixiAI/DeepSeek-R1-0528-AWQ AWQ, 133 downloads
- adamo1139/DeepSeek-R1-0528-AWQ AWQ, 85 downloads
- QuantTrio/DeepSeek-R1-0528-GPTQ-Int4-Int8Mix-Lite GPTQ, int4, 64 downloads
- amd/DeepSeek-R1-0528-MTP-PTPC-FP8 FP8, 51 downloads
- QuantTrio/DeepSeek-R1-0528-GPTQ-Int4-Int8Mix-Compact GPTQ, int4, 36 downloads
- CPU-Hybrid-MoE/DeepSeek-R1-0528-CPU-NUMA4-AMXINT8 FP8, 25 downloads
- xxrjun/DeepSeek-R1-0528-FP4 FP8, 20 downloads
- jasonyux/DeepSeek-R1-0528-AWQ AWQ, 13 downloads
The Hub tags 48 quantised repos as conversions of this model; we read the most downloaded and show the maker's own first.
Run it by API
4 providers serve this model through the Hugging Face router. Prices are US dollars per million tokens. Precision is what the provider reports; a lower precision than the original weights can change answers. Live first, then cheapest output.
| Provider | In | Out | Context | Precision | Tokens/s | First token | Uptime 24h | Tools | JSON |
|---|---|---|---|---|---|---|---|---|---|
| DeepInfra | $0.50 | $2.15 | 160k | fp4 | 29 | 0.5 s | 97.7% | no | yes |
| SiliconFlow | $0.50 | $2.18 | 160k | fp8 | 21 | 1.9 s | 99.7% | yes | yes |
| novita | $0.70 | $2.50 | 160k | not given | 22 | 0.9 s | ? | yes | yes |
| featherless-ai | ? | ? | ? | not given | ? | ? | ? | not given | not given |
Through the router, with a Hugging Face token (the router bills your Hugging Face account). Change the name after the colon to pick another provider.
curl https://router.huggingface.co/v1/chat/completions \
-H "Authorization: Bearer $HF_TOKEN" -H "Content-Type: application/json" \
-d '{"model": "deepseek-ai/DeepSeek-R1-0528:deepinfra", "messages": [{"role": "user", "content": "Hello"}]}'Not built yet: one key for every open model, with BlockRun, and our own tested-quality score per provider. Until then, precision and measured speed are the best guide to a provider here.
Fine-tunes of this model
Repos that declare this model as their base, most downloaded first. Listing is not a recommendation; unreviewed ones have not been checked.
- nvidia/DeepSeek-R1-0528-NVFP4-v217kUnreviewed
- nvidia/DeepSeek-R1-0528-NVFP43.1kUnreviewed
Planned for this page Not built yet
- Try it in private chat
- A link to Solana AI's confidential chat for the models it serves. This model is not listed as served there.
- Fine-tune it
- Fine-tuning a few base models from this page and publishing the result here.
- Creators and earnings
- Pages for people who publish fine-tunes and adapters of this model, and what they earn.
- Fund and bounties
- Bounties for new open models, paid from a public fund. How the fund will work.
- Contribute compute
- How to give compute to open model work, with Akash.
Model data from the Hugging Face Hub 19 Nov 2026, 13:43 UTC. The repo. Found a mistake on this page? Tell us.