- Size
- 30.5B 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
- 28 Jul 2025the date the repo was created
- Context length
- 256k tokensfrom the model's config
- Kind
- Base model (no base model declared on the Hub)Qwen3MoeForCausalLM
- On Hugging Face
- Qwen/Qwen3-30B-A3B-Instruct-2507734k downloads in the last 30 days, 856 likes
Scores
- default setting55.6%±2.930 Aug 2026, log
Epoch's name: qwen3-30b-a3b-instruct-2507. Best closed model on this set: gpt-6.2-astra_max, 96.8% (OpenAI). All 341 results
- default setting62.2%±6.530 Aug 2026, log
Epoch's name: qwen3-30b-a3b-instruct-2507. Best closed model on this set: claude-fable-5_high, 100.0% (Anthropic). All 318 results
- default setting67.2%conceptualreasoning.ai
Epoch's name: qwen3-30b-a3b-instruct-2507. Best closed model on this set: claude-fable-5-2_unknown, 99.2% (Anthropic). All 241 results
- default setting22.4%17.5 to 27.2conceptualreasoning.ai
Epoch's name: qwen3-30b-a3b-instruct-2507. 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: 0.75 GB, worked out from the model's config (Qwen/Qwen3-30B-A3B-Instruct-2507), 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/Qwen3-30B-A3B-Instruct-2507-GGUF (converted by unsloth), 538k downloads. Sizes are the real files on the Hub, all parts added up.
| Format | Size | Memory needed | Get it |
|---|---|---|---|
| UD-IQ3_XXS | 12.0 GBfile size | 13.3 GB12.0 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:UD-IQ3_XXSOllama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:UD-IQ3_XXSDownload hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-UD-IQ3_XXS.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| Q3_K_S | 12.4 GBfile size | 13.6 GB12.4 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q3_K_SOllama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q3_K_SDownload hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-Q3_K_S.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| UD-Q3_K_XL | 12.9 GBfile size | 14.1 GB12.9 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:UD-Q3_K_XLOllama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:UD-Q3_K_XLDownload hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-UD-Q3_K_XL.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| Q3_K_M | 13.7 GBfile size | 15.0 GB13.7 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q3_K_MOllama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q3_K_MDownload hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-Q3_K_M.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| IQ4_XS | 15.3 GBfile size | 16.5 GB15.3 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:IQ4_XSOllama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:IQ4_XSDownload hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-IQ4_XS.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| IQ4_NL | 16.1 GBfile size | 17.4 GB16.1 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:IQ4_NLOllama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:IQ4_NLDownload hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-IQ4_NL.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| Q4_0 | 16.2 GBfile size | 17.4 GB16.2 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_0Ollama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_0Download hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-Q4_0.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| Q4_K_S | 16.3 GBfile size | 17.5 GB16.3 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_SOllama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_SDownload hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-Q4_K_S.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| UD-Q4_K_XL | 16.5 GBfile size | 17.7 GB16.5 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:UD-Q4_K_XLOllama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:UD-Q4_K_XLDownload hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-UD-Q4_K_XL.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| Q4_K_M | 17.3 GBfile size | 18.5 GB17.3 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_MOllama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_K_MDownload hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-Q4_K_M.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| Q4_1 | 17.9 GBfile size | 19.1 GB17.9 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_1Ollama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q4_1Download hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-Q4_1.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| Q5_K_S | 19.6 GBfile size | 20.9 GB19.6 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q5_K_SOllama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q5_K_SDownload hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-Q5_K_S.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| Q5_K_M | 20.2 GBfile size | 21.5 GB20.2 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q5_K_MOllama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q5_K_MDownload hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-Q5_K_M.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| UD-Q5_K_XL | 20.2 GBfile size | 21.5 GB20.2 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:UD-Q5_K_XLOllama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:UD-Q5_K_XLDownload hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-UD-Q5_K_XL.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| Q6_K | 23.4 GBfile size | 24.6 GB23.4 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q6_KOllama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q6_KDownload hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-Q6_K.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| UD-Q6_K_XL | 24.5 GBfile size | 25.8 GB24.5 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:UD-Q6_K_XLOllama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:UD-Q6_K_XLDownload hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-UD-Q6_K_XL.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| Q8_0 | 30.3 GBfile size | 31.5 GB30.3 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q8_0Ollama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:Q8_0Download hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-Q8_0.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| UD-Q8_K_XL | 33.5 GBfile size | 34.8 GB33.5 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:UD-Q8_K_XLOllama ollama run hf.co/unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:UD-Q8_K_XLDownload hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "Qwen3-30B-A3B-Instruct-2507-UD-Q8_K_XL.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
| BF16 | 56.9 GBfile size, 2 parts | 58.1 GB56.9 + 0.75 + 0.50 | Commandsllama.cpp llama-server -hf unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF:BF16Download hf download unsloth/Qwen3-30B-A3B-Instruct-2507-GGUF --include "BF16/Qwen3-30B-A3B-Instruct-2507-BF16-*-of-00002.gguf" --local-dir Qwen3-30B-A3B-Instruct-2507-GGUF |
MLX for Apple silicon
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| MLX 4bitlmstudio-community/Qwen3-30B-A3B-Instruct-2507-MLX-4bit | 16.0 GBfile size | 17.2 GB16.0 + 0.75 + 0.50 | Download page |
| MLX 8bitlmstudio-community/Qwen3-30B-A3B-Instruct-2507-MLX-8bit | 30.2 GBfile size | 31.5 GB30.2 + 0.75 + 0.50 | Download page |
Original weights size from the safetensors header, for vLLM, SGLang, Transformers
| Format | Size | Memory needed | Repo |
|---|---|---|---|
| BF16original weights | 56.9 GBfile size | 58.1 GB56.9 + 0.75 + 0.50 |
hf download Qwen/Qwen3-30B-A3B-Instruct-2507 --local-dir Qwen3-30B-A3B-Instruct-2507Other conversions (15; sizes not read)
- Qwen/Qwen3-30B-A3B-Instruct-2507-FP8 FP8, by the maker, 343k downloads
- stelterlab/Qwen3-30B-A3B-Instruct-2507-AWQ AWQ, 86k downloads
- cyankiwi/Qwen3-30B-A3B-Instruct-2507-AWQ-4bit AWQ, 4bit, 20k downloads
- RedHatAI/Qwen3-30B-A3B-Instruct-2507-quantized.w4a16 Other, w4a16, 19k downloads
- INCModel2/Qwen3-30B-A3B-Instruct-2507-MXFP8-FP8ATTN-CT-AutoRound Other, 2.2k downloads
- NVFP4/Qwen3-30B-A3B-Instruct-2507-FP4 NVFP4, 1.5k downloads
- ELVISIO/Qwen3-30B-A3B-Instruct-2507-AWQ AWQ, 1.4k downloads
- pearl-ai/Qwen3-30B-A3B-Instruct-2507-pearl Other, 991 downloads
- RedHatAI/Qwen3-30B-A3B-Instruct-2507-quantized.w8a8 Other, w8a8, 912 downloads
- JunHowie/Qwen3-30B-A3B-Instruct-2507-GPTQ-Int4 GPTQ, int4, 643 downloads
- a-ivanovitch/Qwen3-30B-A3B-Instruct-2507-AWQ-W4A16 AWQ, w4a16, 384 downloads
- QuantTrio/Qwen3-30B-A3B-Instruct-2507-GPTQ-Int8 GPTQ, int8, 120 downloads
- positron-ai/Qwen_Qwen3-30B-A3B-Instruct-2507-ingest-best-gptq GPTQ, 64 downloads
- llmat/Qwen3-30B-A3B-Instruct-2507-NVFP4 NVFP4, 62 downloads
- s0nakh/Qwen3-30B-A3B-Instruct-2507-kk-ru-GPTQ-4-3bit GPTQ, 3bit, 52 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
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 |
|---|---|---|---|---|---|---|---|---|---|
| DekaLLM | $0.09 | $0.30 | 256k | not given | 74 | 0.5 s | 99.0% | yes | yes |
| Nebius | $0.10 | $0.30 | 256k | fp8 | 36 | 0.3 s | 96.7% | yes | yes |
| SiliconFlow | $0.09 | $0.30 | 256k | fp8 | 37 | 1.0 s | 98.8% | 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/Qwen3-30B-A3B-Instruct-2507:dekallm", "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.
- apple/SimpleSD-30B-instruct30.5B1.3kUnreviewed
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.