- Size
- 403B parameters, 17.0B active (from the repo name)counted from the safetensors header on the Hub
- Licence
- custom, read it, Apache License 2.0Our reading, check the licence: A custom licence we have not read for this model: other makers use the same name for different texts. Read the model's own licence file before you use it.Licence text
- On the Hub since
- 16 Feb 2026the date the repo was created
- Context length
- 256k tokensfrom the model's config
- Kind
- Base model (no base model declared on the Hub)Qwen3_5MoeForConditionalGeneration
- On Hugging Face
- Qwen/Qwen3.5-397B-A17B677k downloads in the last 30 days, 1.6k likes
Scores
Epoch's name: qwen3.5-397b-a17b. Best closed model on this set: claude-fable-5-2_max, 95.1% (Anthropic). All 127 results
Epoch's name: qwen3.5-397b-a17b. Best closed model on this set: gpt-6.2-astra_max, 96.8% (OpenAI). All 341 results
- effort none18.0%±3.927 Aug 2026, log
Epoch's name: qwen3.5-397b-a17b. Best closed model on this set: gpt-6.2-astra_max, 95.0% (OpenAI). All 148 results
Epoch's name: qwen3.5-397b-a17b. Best closed model on this set: claude-fable-5_high, 100.0% (Anthropic). All 318 results
- default setting24.9%via Epoch AI (error bar left out)
Epoch's name: qwen3.5-397b-a17b. Best closed model on this set: gemini-4-carbon_unknown, 84.2% (Google DeepMind). All 66 results
- effort none0.9%via Epoch AI
Epoch's name: qwen3.5-397b-a17b. Best closed model on this set: gemini-4-carbon_high, 36.0% (Google DeepMind). All 271 results
- default setting87.5%conceptualreasoning.ai
Epoch's name: qwen3.5-397b-a17b. Best closed model on this set: claude-fable-5-2_unknown, 99.2% (Anthropic). All 241 results
- default setting37.9%33.5 to 42.3conceptualreasoning.ai
Epoch's name: qwen3.5-397b-a17b. Best closed model on this set: claude-fable-5-2_xhigh, 70.6% (Anthropic). All 203 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: 0.23 GB, worked out from the model's config (Qwen/Qwen3.5-397B-A17B), 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/Qwen3.5-397B-A17B-GGUF (converted by unsloth), 26k downloads. Sizes are the real files on the Hub, all parts added up. For image input you also need the projector file mmproj-F16.gguf (0.86 GB), not counted below.
| Format | Size | Memory needed | Get it |
|---|---|---|---|
| UD-IQ3_XXS | 130.7 GBfile size, 4 parts | 131.4 GB130.7 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:UD-IQ3_XXSDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "UD-IQ3_XXS/Qwen3.5-397B-A17B-UD-IQ3_XXS-*-of-00004.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| UD-IQ3_S | 136.3 GBfile size, 4 parts | 137.1 GB136.3 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:UD-IQ3_SDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "UD-IQ3_S/Qwen3.5-397B-A17B-UD-IQ3_S-*-of-00004.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| Q3_K_S | 153.0 GBfile size, 5 parts | 153.8 GB153.0 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:Q3_K_SDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "Q3_K_S/Qwen3.5-397B-A17B-Q3_K_S-*-of-00005.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| UD-Q3_K_S | 153.0 GBfile size, 5 parts | 153.8 GB153.0 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:UD-Q3_K_SDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "UD-Q3_K_S/Qwen3.5-397B-A17B-UD-Q3_K_S-*-of-00005.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| Q3_K_M | 165.2 GBfile size, 5 parts | 166.0 GB165.2 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:Q3_K_MDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "Q3_K_M/Qwen3.5-397B-A17B-Q3_K_M-*-of-00005.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| UD-Q3_K_M | 165.2 GBfile size, 5 parts | 166.0 GB165.2 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:UD-Q3_K_MDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "UD-Q3_K_M/Qwen3.5-397B-A17B-UD-Q3_K_M-*-of-00005.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| UD-Q3_K_XL | 166.5 GBfile size, 5 parts | 167.2 GB166.5 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:UD-Q3_K_XLDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "UD-Q3_K_XL/Qwen3.5-397B-A17B-UD-Q3_K_XL-*-of-00005.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| UD-IQ4_XS | 176.7 GBfile size, 5 parts | 177.4 GB176.7 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:UD-IQ4_XSDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "UD-IQ4_XS/Qwen3.5-397B-A17B-UD-IQ4_XS-*-of-00005.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| UD-IQ4_NL | 180.5 GBfile size, 5 parts | 181.2 GB180.5 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:UD-IQ4_NLDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "UD-IQ4_NL/Qwen3.5-397B-A17B-UD-IQ4_NL-*-of-00005.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| Q4_K_S | 212.3 GBfile size, 6 parts | 213.1 GB212.3 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:Q4_K_SDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "Q4_K_S/Qwen3.5-397B-A17B-Q4_K_S-*-of-00006.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| UD-Q4_K_S | 212.3 GBfile size, 6 parts | 213.1 GB212.3 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:UD-Q4_K_SDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "UD-Q4_K_S/Qwen3.5-397B-A17B-UD-Q4_K_S-*-of-00006.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| MXFP4_MOE | 221.1 GBfile size, 6 parts | 221.8 GB221.1 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:MXFP4_MOEDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "MXFP4_MOE/Qwen3.5-397B-A17B-MXFP4_MOE-*-of-00006.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| Q4_K_M | 227.3 GBfile size, 6 parts | 228.1 GB227.3 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:Q4_K_MDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "Q4_K_M/Qwen3.5-397B-A17B-Q4_K_M-*-of-00006.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| UD-Q4_K_M | 227.3 GBfile size, 6 parts | 228.1 GB227.3 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:UD-Q4_K_MDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "UD-Q4_K_M/Qwen3.5-397B-A17B-UD-Q4_K_M-*-of-00006.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| UD-Q4_K_XL | 228.4 GBfile size, 6 parts | 229.2 GB228.4 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:UD-Q4_K_XLDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "UD-Q4_K_XL/Qwen3.5-397B-A17B-UD-Q4_K_XL-*-of-00006.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| Q5_K_S | 257.6 GBfile size, 7 parts | 258.3 GB257.6 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:Q5_K_SDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "Q5_K_S/Qwen3.5-397B-A17B-Q5_K_S-*-of-00007.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| UD-Q5_K_S | 257.6 GBfile size, 7 parts | 258.3 GB257.6 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:UD-Q5_K_SDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "UD-Q5_K_S/Qwen3.5-397B-A17B-UD-Q5_K_S-*-of-00007.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| Q5_K_M | 273.5 GBfile size, 8 parts | 274.2 GB273.5 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:Q5_K_MDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "Q5_K_M/Qwen3.5-397B-A17B-Q5_K_M-*-of-00008.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| UD-Q5_K_M | 273.5 GBfile size, 8 parts | 274.2 GB273.5 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:UD-Q5_K_MDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "UD-Q5_K_M/Qwen3.5-397B-A17B-UD-Q5_K_M-*-of-00008.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| UD-Q5_K_XL | 274.6 GBfile size, 8 parts | 275.3 GB274.6 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:UD-Q5_K_XLDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "UD-Q5_K_XL/Qwen3.5-397B-A17B-UD-Q5_K_XL-*-of-00008.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| Q6_K | 304.2 GBfile size, 8 parts | 304.9 GB304.2 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:Q6_KDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "Q6_K/Qwen3.5-397B-A17B-Q6_K-*-of-00008.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| UD-Q6_K | 304.2 GBfile size, 8 parts | 304.9 GB304.2 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:UD-Q6_KDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "UD-Q6_K/Qwen3.5-397B-A17B-UD-Q6_K-*-of-00008.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| UD-Q6_K_XL | 337.4 GBfile size, 9 parts | 338.2 GB337.4 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:UD-Q6_K_XLDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "UD-Q6_K_XL/Qwen3.5-397B-A17B-UD-Q6_K_XL-*-of-00009.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| Q8_0 | 392.6 GBfile size, 10 parts | 393.3 GB392.6 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:Q8_0Download hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "Q8_0/Qwen3.5-397B-A17B-Q8_0-*-of-00010.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| UD-Q8_K_XL | 398.4 GBfile size, 10 parts | 399.1 GB398.4 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:UD-Q8_K_XLDownload hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "UD-Q8_K_XL/Qwen3.5-397B-A17B-UD-Q8_K_XL-*-of-00010.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
| BF16 | 738.5 GBfile size, 17 parts | 739.2 GB738.5 + 0.23 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:BF16Download hf download unsloth/Qwen3.5-397B-A17B-GGUF --include "BF16/Qwen3.5-397B-A17B-BF16-*-of-00017.gguf" --local-dir Qwen3.5-397B-A17B-GGUF |
MLX for Apple silicon
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| MLX nvfp4mlx-community/Qwen3.5-397B-A17B-nvfp4 | 207.7 GBfile size | 208.4 GB207.7 + 0.23 + 0.50 | Download page |
| MLX 8bitlmstudio-community/Qwen3.5-397B-A17B-MLX-8bit | 393.0 GBfile size | 393.8 GB393.0 + 0.23 + 0.50 | Download page |
Original weights size from the safetensors header, for vLLM, SGLang, Transformers
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| BF16original weights | 751.4 GBfile size | 752.1 GB751.4 + 0.23 + 0.50 |
hf download Qwen/Qwen3.5-397B-A17B --local-dir Qwen3.5-397B-A17BOther conversions (19; sizes not read)
- Qwen/Qwen3.5-397B-A17B-FP8 FP8, by the maker, 104k downloads
- Qwen/Qwen3.5-397B-A17B-GPTQ-Int4 GPTQ, int4, by the maker, 4.5k downloads
- nvidia/Qwen3.5-397B-A17B-NVFP4-V2 NVFP4, 36k downloads
- nvidia/Qwen3.5-397B-A17B-NVFP4 NVFP4, 8.5k downloads
- amd/Qwen3.5-397B-A17B-MXFP4-AttnFP8-V2 Other, mxfp4, 4.1k downloads
- cyankiwi/Qwen3.5-397B-A17B-AWQ-4bit AWQ, 4bit, 1.7k downloads
- QuantTrio/Qwen3.5-397B-A17B-AWQ AWQ, 900 downloads
- RedHatAI/Qwen3.5-397B-A17B-FP8-dynamic FP8, 615 downloads
- Intel/Qwen3.5-397B-A17B-int4-AutoRound Other, int4, 559 downloads
- intel-ai/Qwen3.5-397B-A17B-MXFP4-Mixed-CT-AutoRound Other, mxfp4, 270 downloads
- amd/Qwen3.5-397B-A17B-NVFP4 NVFP4, 239 downloads
- amd/Qwen3.5-397B-A17B-MXFP4-AttnFP8-V3 Other, mxfp4, 227 downloads
- AxionML/Qwen3.5-397B-A17B-NVFP4 NVFP4, 168 downloads
- Cialtion/SimpleTool-Qwen3.5-397B-A17B-8Head-5pct-FP8 FP8, 78 downloads
- mratsim/Qwen3.5-397B-A17B-EXL3 EXL3, 70 downloads
- thehosy/Qwen3.5-397B-A17B-MXFP4-MTP-BF16 FP8, mxfp4, 27 downloads
- NeuroSenko/Qwen3.5-397B-A17B-exl3 EXL3, 9 downloads
- cpral/Qwen3.5-397B-A17B-exl3 EXL3, 8 downloads
- MikeRoz/Qwen3.5-397B-A17B-exl3 EXL3, 3 downloads
The Hub tags 81 quantised repos as conversions of this model; we read the most downloaded and show the maker's own first.
Run it by API
6 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 | $0.45 | $3.00 | 256k | fp8 | 49 | 1.0 s | 98.2% | yes | yes |
| Novita | $0.60 | $3.60 | 256k | not given | 38 | 1.2 s | 100.0% | yes | no |
| Parasail | $0.50 | $3.60 | 256k | fp8 | 61 | 0.7 s | 99.0% | yes | yes |
| ovhcloud | $0.71 | $4.25 | 256k | not given | 39 | 0.8 s | ? | yes | yes |
| Venice | $0.75 | $4.50 | 125k | not given | 47 | 0.6 s | 98.5% | 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": "Qwen/Qwen3.5-397B-A17B: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.
- tiny-random/qwen3.5-moe5M211kUnreviewed
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.