- Size
- 9.7B parameterscounted 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
- 27 Feb 2026the date the repo was created
- Context length
- 256k tokensfrom the model's config
- Fine-tune
- of Qwen/Qwen3.5-9B-BaseQwen3_5ForConditionalGeneration
- On Hugging Face
- Qwen/Qwen3.5-9B6.4M downloads in the last 30 days, 2.4k likes
Scores
Epoch's name: qwen3.5-9B. Best closed model on this set: gpt-6.2-astra_max, 96.8% (OpenAI). All 341 results
Epoch's name: qwen3.5-9B. Best closed model on this set: claude-fable-5_high, 100.0% (Anthropic). All 318 results
- effort none0.6%via Epoch AI
- default setting0.3%via Epoch AI
Epoch's name: qwen3.5-9B. Best closed model on this set: gemini-4-carbon_high, 36.0% (Google DeepMind). All 271 results
- default setting71.2%conceptualreasoning.ai
Epoch's name: qwen3.5-9B. Best closed model on this set: claude-fable-5-2_unknown, 99.2% (Anthropic). All 241 results
- default setting24.5%19.6 to 29.3conceptualreasoning.ai
Epoch's name: qwen3.5-9B. Best closed model on this set: claude-fable-5-2_xhigh, 70.6% (Anthropic). All 203 results
- Agent: little-coder9.2%±2.42 Mar 2026, tbench.ai
Epoch's name: qwen3.5-9B. Best closed model on this set: gpt-5.5_unknown, 84.7% (OpenAI). All 202 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: 0.25 GB, worked out from the model's config (Qwen/Qwen3.5-9B), 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/Qwen3.5-9B-GGUF (converted by unsloth), 2.0M 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 | 3.7 GBfile size | 4.5 GB3.7 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:UD-IQ3_XXSOllama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:UD-IQ3_XXSDownload hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-UD-IQ3_XXS.gguf" --local-dir Qwen3.5-9B-GGUF |
| Q3_K_S | 4.0 GBfile size | 4.8 GB4.0 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:Q3_K_SOllama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:Q3_K_SDownload hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-Q3_K_S.gguf" --local-dir Qwen3.5-9B-GGUF |
| Q3_K_M | 4.4 GBfile size | 5.1 GB4.4 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:Q3_K_MOllama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:Q3_K_MDownload hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-Q3_K_M.gguf" --local-dir Qwen3.5-9B-GGUF |
| UD-Q3_K_XL | 4.7 GBfile size | 5.5 GB4.7 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:UD-Q3_K_XLOllama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:UD-Q3_K_XLDownload hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-UD-Q3_K_XL.gguf" --local-dir Qwen3.5-9B-GGUF |
| IQ4_XS | 4.8 GBfile size | 5.6 GB4.8 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:IQ4_XSOllama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:IQ4_XSDownload hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-IQ4_XS.gguf" --local-dir Qwen3.5-9B-GGUF |
| IQ4_NL | 5.0 GBfile size | 5.8 GB5.0 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:IQ4_NLOllama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:IQ4_NLDownload hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-IQ4_NL.gguf" --local-dir Qwen3.5-9B-GGUF |
| Q4_0 | 5.0 GBfile size | 5.8 GB5.0 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:Q4_0Ollama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:Q4_0Download hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-Q4_0.gguf" --local-dir Qwen3.5-9B-GGUF |
| Q4_K_S | 5.0 GBfile size | 5.8 GB5.0 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:Q4_K_SOllama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:Q4_K_SDownload hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-Q4_K_S.gguf" --local-dir Qwen3.5-9B-GGUF |
| Q4_K_M | 5.3 GBfile size | 6.0 GB5.3 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:Q4_K_MOllama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:Q4_K_MDownload hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-Q4_K_M.gguf" --local-dir Qwen3.5-9B-GGUF |
| Q4_1 | 5.4 GBfile size | 6.2 GB5.4 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:Q4_1Ollama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:Q4_1Download hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-Q4_1.gguf" --local-dir Qwen3.5-9B-GGUF |
| UD-Q4_K_XL | 5.6 GBfile size | 6.3 GB5.6 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:UD-Q4_K_XLOllama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:UD-Q4_K_XLDownload hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-UD-Q4_K_XL.gguf" --local-dir Qwen3.5-9B-GGUF |
| Q5_K_S | 5.9 GBfile size | 6.7 GB5.9 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:Q5_K_SOllama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:Q5_K_SDownload hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-Q5_K_S.gguf" --local-dir Qwen3.5-9B-GGUF |
| Q5_K_M | 6.1 GBfile size | 6.9 GB6.1 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:Q5_K_MOllama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:Q5_K_MDownload hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-Q5_K_M.gguf" --local-dir Qwen3.5-9B-GGUF |
| UD-Q5_K_XL | 6.3 GBfile size | 7.0 GB6.3 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:UD-Q5_K_XLOllama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:UD-Q5_K_XLDownload hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-UD-Q5_K_XL.gguf" --local-dir Qwen3.5-9B-GGUF |
| Q6_K | 6.9 GBfile size | 7.7 GB6.9 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:Q6_KOllama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:Q6_KDownload hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-Q6_K.gguf" --local-dir Qwen3.5-9B-GGUF |
| UD-Q6_K_XL | 8.2 GBfile size | 8.9 GB8.2 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:UD-Q6_K_XLOllama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:UD-Q6_K_XLDownload hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-UD-Q6_K_XL.gguf" --local-dir Qwen3.5-9B-GGUF |
| Q8_0 | 8.9 GBfile size | 9.6 GB8.9 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:Q8_0Ollama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:Q8_0Download hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-Q8_0.gguf" --local-dir Qwen3.5-9B-GGUF |
| UD-Q8_K_XL | 12.1 GBfile size | 12.8 GB12.1 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:UD-Q8_K_XLOllama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:UD-Q8_K_XLDownload hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-UD-Q8_K_XL.gguf" --local-dir Qwen3.5-9B-GGUF |
| BF16 | 16.7 GBfile size | 17.4 GB16.7 + 0.25 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3.5-9B-GGUF:BF16Ollama ollama run hf.co/unsloth/Qwen3.5-9B-GGUF:BF16Download hf download unsloth/Qwen3.5-9B-GGUF --include "Qwen3.5-9B-BF16.gguf" --local-dir Qwen3.5-9B-GGUF |
MLX for Apple silicon
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| MLX 4bitlmstudio-community/Qwen3.5-9B-MLX-4bit | 5.5 GBfile size | 6.3 GB5.5 + 0.25 + 0.50 | Download page |
| MLX 8bitlmstudio-community/Qwen3.5-9B-MLX-8bit | 9.7 GBfile size | 10.5 GB9.7 + 0.25 + 0.50 | Download page |
Original weights size from the safetensors header, for vLLM, SGLang, Transformers
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| BF16original weights | 18.0 GBfile size | 18.7 GB18.0 + 0.25 + 0.50 |
hf download Qwen/Qwen3.5-9B --local-dir Qwen3.5-9BOther conversions (13; sizes not read)
- RedHatAI/Qwen3.5-9B-FP8-dynamic FP8, 568k downloads
- cyankiwi/Qwen3.5-9B-AWQ-4bit AWQ, 4bit, 186k downloads
- AxionML/Qwen3.5-9B-NVFP4 NVFP4, 173k downloads
- QuantTrio/Qwen3.5-9B-AWQ AWQ, 26k downloads
- lovedheart/Qwen3.5-9B-FP8 FP8, 14k downloads
- RedHatAI/Qwen3.5-9B-quantized.w4a16 Other, w4a16, 14k downloads
- nicklas373/Qwen3.5-9B-AWQ AWQ, 13k downloads
- cyankiwi/Qwen3.5-9B-AWQ-BF16-INT8 AWQ, bf16, 5.4k downloads
- Freesolo-Co/Qwen3.5-9B-FP8 FP8, 4.2k downloads
- Intel/Qwen3.5-9B-int4-AutoRound Other, int4, 2.0k downloads
- litert-community/Qwen3.5-9B Other, 1.5k downloads
- ruwwww/qwen3.5-9b-ninfer Other, 1.4k downloads
- RedHatAI/Qwen3.5-9B-NVFP4 NVFP4, 1.3k 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
7 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 | $0.10 | $0.15 | 256k | bf16 | 72 | 0.6 s | 99.3% | yes | yes |
| SiliconFlow | $0.10 | $0.15 | 256k | fp8 | 23 | 1.6 s | 99.3% | yes | yes |
| ovhcloudserves cyankiwi/Qwen3.5-9B-AWQ-4bit | $0.12 | $0.18 | 256k | not given | 111 | 0.4 s | ? | yes | yes |
| PrimeIntellect | $0.14 | $0.21 | 256k | bf16 | 108 | 0.2 s | 100.0% | yes | yes |
| Parasail | $0.10 | $0.25 | 256k | bf16 | 80 | 0.6 s | 99.9% | no | yes |
| Together | $0.17 | $0.25 | 256k | not given | 91 | 0.3 s | 99.6% | 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/Qwen3.5-9B: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.
- unsloth/Qwen3.5-9B9.7B112kUnreviewed
- prithivMLmods/Qwen3.5-9B-Uncensored-ComfyUI43kUnreviewed
- z-lab/Qwen3.5-9B-DFlash42kUnreviewed
- black-forest-labs/flux-3-image-klein-pup9.4B40kUnreviewed
- Cloudflare/clef-flash9.4B33kUnreviewed
- XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B9.4B31kReviewed
- prism-ml/Ternary-Bonsai-2-9B-mlx-2bit16kUnreviewed
- DICPlay/RusMind-Sokol-4.79.0B13kUnreviewed
- huihui-ai/Huihui-Qwen3.5-9B-abliterated9.7B7.8kUnreviewed
- inclusionAI/UI-Venus-2-9B6.3kUnreviewed
- empero-ai/Qwen3.8-9B-Distill9.7B5.4kUnreviewed
- microsoft/Fara1.5-9B9.4B4.0kUnreviewed
- morriszjm/Tacit-9B9.4B3.8kUnreviewed
- empero-ai/Qwythos-9B-Claude-Mythos-5-1M9.4B3.6kUnreviewed
- lukey03/Qwen3.5-9B-abliterated9.0B2.4kUnreviewed
- tencent/UI-Mate-9B9.4B2.1kUnreviewed
- inclusionAI/ZwZ2-9B9.4B1.4kUnreviewed
- 169Pi/Alpie_seed9.4B1.2kUnreviewed
- apple/LensVLM-9B9.4B900Unreviewed
- OrionLLM/OxCoder-9B9.4B624Unreviewed
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.