- Size
- 36.0B parameters, 3.0B 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
- 24 Feb 2026the date the repo was created
- Context length
- 256k tokensfrom the model's config
- Fine-tune
- of Qwen/Qwen3.5-35B-A3B-BaseQwen3_5MoeForConditionalGeneration
- On Hugging Face
- Qwen/Qwen3.5-35B-A3B1.4M downloads in the last 30 days, 1.6k likes
Scores
Epoch's name: qwen3.5-35b-a3b. Best closed model on this set: gpt-6.2-astra_max, 96.8% (OpenAI). All 341 results
Epoch's name: qwen3.5-35b-a3b. Best closed model on this set: claude-fable-5_high, 100.0% (Anthropic). All 318 results
- effort none0.6%via Epoch AI
Epoch's name: qwen3.5-35b-a3b. Best closed model on this set: gemini-4-carbon_high, 36.0% (Google DeepMind). All 271 results
- default setting80.0%conceptualreasoning.ai
Epoch's name: qwen3.5-35b-a3b. Best closed model on this set: claude-fable-5-2_unknown, 99.2% (Anthropic). All 241 results
- default setting29.5%24.8 to 34.2conceptualreasoning.ai
Epoch's name: qwen3.5-35b-a3b. 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: 0.16 GB, worked out from the model's config (Qwen/Qwen3.5-35B-A3B), 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/Qwen3.5-35B-A3B-GGUF (converted by unsloth), 389k 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 (0.84 GB), not counted below.
| Format | Size | Memory needed | Get it |
|---|---|---|---|
| UD-IQ3_XXS | 12.2 GBfile size | 12.8 GB12.2 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:UD-IQ3_XXSOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:UD-IQ3_XXSDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-UD-IQ3_XXS.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| UD-IQ3_S | 12.7 GBfile size | 13.3 GB12.7 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:UD-IQ3_SOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:UD-IQ3_SDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-UD-IQ3_S.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| Q3_K_S | 14.2 GBfile size | 14.9 GB14.2 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:Q3_K_SOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:Q3_K_SDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-Q3_K_S.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| Q3_K_M | 15.2 GBfile size | 15.9 GB15.2 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:Q3_K_MOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:Q3_K_MDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-Q3_K_M.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| UD-Q3_K_XL | 15.5 GBfile size | 16.1 GB15.5 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q3_K_XLOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q3_K_XLDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-UD-Q3_K_XL.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| UD-IQ4_XS | 16.3 GBfile size | 16.9 GB16.3 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:UD-IQ4_XSOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:UD-IQ4_XSDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-UD-IQ4_XS.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| UD-IQ4_NL | 16.6 GBfile size | 17.3 GB16.6 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:UD-IQ4_NLOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:UD-IQ4_NLDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-UD-IQ4_NL.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| UD-Q4_K_L | 18.8 GBfile size | 19.5 GB18.8 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q4_K_LOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q4_K_LDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-UD-Q4_K_L.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| Q4_K_S | 19.3 GBfile size | 19.9 GB19.3 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:Q4_K_SOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:Q4_K_SDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-Q4_K_S.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| MXFP4_MOE | 20.1 GBfile size | 20.8 GB20.1 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:MXFP4_MOEOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:MXFP4_MOEDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-MXFP4_MOE.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| Q4_K_M | 20.5 GBfile size | 21.2 GB20.5 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:Q4_K_MOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:Q4_K_MDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-Q4_K_M.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| UD-Q4_K_XL | 20.7 GBfile size | 21.4 GB20.7 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q4_K_XLOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q4_K_XLDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-UD-Q4_K_XL.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| Q5_K_S | 23.1 GBfile size | 23.8 GB23.1 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:Q5_K_SOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:Q5_K_SDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-Q5_K_S.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| Q5_K_M | 24.4 GBfile size | 25.1 GB24.4 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:Q5_K_MOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:Q5_K_MDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-Q5_K_M.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| UD-Q5_K_XL | 24.6 GBfile size | 25.2 GB24.6 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q5_K_XLOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q5_K_XLDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-UD-Q5_K_XL.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| UD-Q6_K_S | 26.6 GBfile size | 27.2 GB26.6 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q6_K_SOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q6_K_SDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-UD-Q6_K_S.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| Q6_K | 26.9 GBfile size | 27.5 GB26.9 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:Q6_KOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:Q6_KDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-Q6_K.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| UD-Q6_K_XL | 29.9 GBfile size | 30.5 GB29.9 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q6_K_XLOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q6_K_XLDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-UD-Q6_K_XL.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| Q8_0 | 34.4 GBfile size | 35.0 GB34.4 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:Q8_0Ollama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:Q8_0Download hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-Q8_0.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| UD-Q8_K_XL | 45.3 GBfile size | 46.0 GB45.3 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q8_K_XLOllama ollama run hf.co/unsloth/Qwen3.5-35B-A3B-GGUF:UD-Q8_K_XLDownload hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "Qwen3.5-35B-A3B-UD-Q8_K_XL.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
| BF16 | 64.6 GBfile size, 2 parts | 65.3 GB64.6 + 0.16 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-35B-A3B-GGUF:BF16Download hf download unsloth/Qwen3.5-35B-A3B-GGUF --include "BF16/Qwen3.5-35B-A3B-BF16-*-of-00002.gguf" --local-dir Qwen3.5-35B-A3B-GGUF |
MLX for Apple silicon
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| MLX 3bitandrevp/Qwen3.5-35B-A3B-MLX-VLM-3bit | 15.0 GBfile size | 15.6 GB15.0 + 0.16 + 0.50 | Download page |
| MLX 4bitmlx-community/Qwen3.5-35B-A3B-OptiQ-4bit | 21.9 GBfile size | 22.6 GB21.9 + 0.16 + 0.50 | Download page |
Original weights size from the safetensors header, for vLLM, SGLang, Transformers
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| BF16original weights | 67.0 GBfile size | 67.6 GB67.0 + 0.16 + 0.50 |
hf download Qwen/Qwen3.5-35B-A3B --local-dir Qwen3.5-35B-A3BOther conversions (16; sizes not read)
- Qwen/Qwen3.5-35B-A3B-FP8 FP8, by the maker, 785k downloads
- Qwen/Qwen3.5-35B-A3B-GPTQ-Int4 GPTQ, int4, by the maker, 318k downloads
- QuantTrio/Qwen3.5-35B-A3B-AWQ AWQ, 33k downloads
- AxionML/Qwen3.5-35B-A3B-NVFP4 NVFP4, 7.2k downloads
- txn545/Qwen3.5-35B-A3B-NVFP4 NVFP4, 7.1k downloads
- cyankiwi/Qwen3.5-35B-A3B-AWQ-4bit AWQ, 4bit, 5.7k downloads
- litert-community/Qwen3.5-35B-A3B Other, 3.1k downloads
- SocialLocalMobile/Qwen3.5-35B-A3B-HQQ-INT4 Other, int4, 594 downloads
- taobao-mnn/Qwen3.5-35B-A3B-MNN Other, 581 downloads
- RedHatAI/Qwen3.5-35B-A3B-FP8-dynamic FP8, 368 downloads
- Sehyo/Qwen3.5-35B-A3B-NVFP4 NVFP4, 362 downloads
- apolo13x/Qwen3.5-35B-A3B-quantized.w4a16 GPTQ, w4a16, 351 downloads
- Intel/Qwen3.5-35B-A3B-int4-AutoRound Other, int4, 295 downloads
- turboderp/Qwen3.5-35B-A3B-exl3 EXL3, 246 downloads
- RedHatAI/Qwen3.5-35B-A3B-quantized.w8a8 GPTQ, w8a8, 212 downloads
- ig1/Qwen3.5-35B-A3B-NVFP4 NVFP4, 161 downloads
The Hub tags 100 or more quantised repos as conversions of this model; we read the most downloaded and show the maker's own first.
Run it by API
5 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 | $0.14 | $1.00 | 256k | not given | 120 | 0.4 s | ? | yes | yes |
| Parasail | $0.15 | $1.00 | 256k | fp8 | 53 | 0.9 s | 99.9% | yes | yes |
| Venice | $0.15 | $1.00 | 250k | not given | 135 | 1.1 s | 99.9% | yes | yes |
| SiliconFlow | $0.24 | $1.80 | 256k | fp8 | 71 | 3.2 s | 96.2% | no | yes |
| novita | $0.25 | $2.00 | 256k | not given | 72 | 1.1 s | ? | yes | no |
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/Qwen3.5-35B-A3B: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/Ising-Calibration-1-35B-A3B1.9kUnreviewed
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.