Skip to content
Weights Not Walls
Menu

Qwen

Qwen3.8-27B

Reviewedaccepts images
Size
27.8B parameterscounted from the safetensors header on the Hub
Licence
unknown, Apache License 2.0Our reading, check the licence: You can use, change and redistribute the model, including commercially. Keep the licence and any notices, and say which files you changed; you do not have to publish your changes. It includes a patent licence that ends for anyone who sues over patents in the work.
On the Hub since
5 Aug 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_5ForConditionalGeneration
On Hugging Face
Qwen/Qwen3.8-27B6.6M downloads in the last 30 days, 21k likes

Scores

19 independent results on 10 eval sets
APEX-AgentsReported

Epoch's name: qwen3.8-27b. Best closed model on this set: gemini-4-carbon_unknown, 84.2% (Google DeepMind). All 66 results

ARC-AGIReported

Epoch's name: qwen3.8-27b. Best closed model on this set: claude-fable-5_max, 98.5% (Anthropic). All 318 results

ARC-AGI-2Reported

Epoch's name: qwen3.8-27b. Best closed model on this set: gpt-6-astra_max, 95.0% (OpenAI). All 299 results

CritPtReported

Epoch's name: qwen3.8-27b. Best closed model on this set: gemini-4-carbon_high, 36.0% (Google DeepMind). All 271 results

DTBenchReported

Epoch's name: qwen3.8-27b. Best closed model on this set: claude-fable-5-2_unknown, 99.2% (Anthropic). All 241 results

GeoBenchReported

Epoch's name: qwen3.8-27b. Best closed model on this set: claude-fable-5_high, 89.0% (Anthropic). All 51 results

LMCAReported

Epoch's name: qwen3.8-27b. Best closed model on this set: claude-fable-5-2_xhigh, 70.6% (Anthropic). All 203 results

ProofBenchReported

Epoch's name: qwen3.8-27b. Best closed model on this set: claude-fable-5-1_max, 100.0% (Anthropic). All 85 results

SimpleBenchReported

Epoch's name: qwen3.8-27b. Best closed model on this set: claude-fable-5-2_unknown, 94.9% (Anthropic). All 125 results

Epoch's name: qwen3.8-27b. Best closed model on this set: claude-fable-5_high, 95.0% (Anthropic). All 29 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

file sizes read from the Hub 19 Nov 2026

Memory needed = weights + context cache + 0.50 GB for the runtime. Context cache at 8k tokens: 0.50 GB, worked out from the model's config (Qwen/Qwen3.8-27B), 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/Qwen3.8-27B-GGUF (converted by unsloth), 7.1M 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.

FormatSizeMemory neededGet it
UD-IQ3_XXS10.2 GBfile size11.2 GB10.2 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-IQ3_XXS
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:UD-IQ3_XXS
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-UD-IQ3_XXS.gguf" --local-dir Qwen3.8-27B-GGUF
UD-IQ3_S11.2 GBfile size12.2 GB11.2 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-IQ3_S
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:UD-IQ3_S
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-UD-IQ3_S.gguf" --local-dir Qwen3.8-27B-GGUF
UD-Q3_K_XL12.2 GBfile size13.2 GB12.2 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-Q3_K_XL
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:UD-Q3_K_XL
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-UD-Q3_K_XL.gguf" --local-dir Qwen3.8-27B-GGUF
UD-IQ4_XS13.4 GBfile size14.4 GB13.4 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-IQ4_XS
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:UD-IQ4_XS
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-UD-IQ4_XS.gguf" --local-dir Qwen3.8-27B-GGUF
UD-Q4_K_S14.3 GBfile size15.3 GB14.3 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-Q4_K_S
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:UD-Q4_K_S
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-UD-Q4_K_S.gguf" --local-dir Qwen3.8-27B-GGUF
Q4_015.0 GBfile size16.0 GB15.0 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:Q4_0
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:Q4_0
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-Q4_0.gguf" --local-dir Qwen3.8-27B-GGUF
UD-Q4_K_M15.4 GBfile size16.4 GB15.4 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-Q4_K_M
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:UD-Q4_K_M
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-UD-Q4_K_M.gguf" --local-dir Qwen3.8-27B-GGUF
Q4_116.3 GBfile size17.3 GB16.3 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:Q4_1
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:Q4_1
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-Q4_1.gguf" --local-dir Qwen3.8-27B-GGUF
UD-Q4_K_XL16.4 GBfile size17.4 GB16.4 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-Q4_K_XL
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:UD-Q4_K_XL
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-UD-Q4_K_XL.gguf" --local-dir Qwen3.8-27B-GGUF
UD-Q5_K_S17.5 GBfile size18.5 GB17.5 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-Q5_K_S
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:UD-Q5_K_S
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-UD-Q5_K_S.gguf" --local-dir Qwen3.8-27B-GGUF
UD-Q5_K_M18.5 GBfile size19.5 GB18.5 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-Q5_K_M
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:UD-Q5_K_M
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-UD-Q5_K_M.gguf" --local-dir Qwen3.8-27B-GGUF
UD-Q5_K_XL19.5 GBfile size20.5 GB19.5 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-Q5_K_XL
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:UD-Q5_K_XL
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-UD-Q5_K_XL.gguf" --local-dir Qwen3.8-27B-GGUF
UD-Q6_K20.5 GBfile size21.5 GB20.5 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-Q6_K
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:UD-Q6_K
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-UD-Q6_K.gguf" --local-dir Qwen3.8-27B-GGUF
UD-Q6_K_M21.5 GBfile size22.5 GB21.5 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-Q6_K_M
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:UD-Q6_K_M
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-UD-Q6_K_M.gguf" --local-dir Qwen3.8-27B-GGUF
UD-Q6_K_L22.5 GBfile size23.5 GB22.5 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-Q6_K_L
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:UD-Q6_K_L
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-UD-Q6_K_L.gguf" --local-dir Qwen3.8-27B-GGUF
UD-Q6_K_XL23.6 GBfile size24.6 GB23.6 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-Q6_K_XL
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:UD-Q6_K_XL
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-UD-Q6_K_XL.gguf" --local-dir Qwen3.8-27B-GGUF
UD-Q8_K_L26.1 GBfile size27.1 GB26.1 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-Q8_K_L
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:UD-Q8_K_L
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-UD-Q8_K_L.gguf" --local-dir Qwen3.8-27B-GGUF
Q8_027.1 GBfile size28.1 GB27.1 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:Q8_0
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:Q8_0
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-Q8_0.gguf" --local-dir Qwen3.8-27B-GGUF
UD-Q8_K_XL29.3 GBfile size30.3 GB29.3 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:UD-Q8_K_XL
Ollama
ollama run hf.co/unsloth/Qwen3.8-27B-GGUF:UD-Q8_K_XL
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "Qwen3.8-27B-UD-Q8_K_XL.gguf" --local-dir Qwen3.8-27B-GGUF
BF1650.9 GBfile size, 2 parts51.9 GB50.9 + 0.50 + 0.50
Commands
llama.cpp
llama-server -hf unsloth/Qwen3.8-27B-GGUF:BF16
Download
hf download unsloth/Qwen3.8-27B-GGUF --include "BF16/Qwen3.8-27B-BF16-*-of-00002.gguf" --local-dir Qwen3.8-27B-GGUF

MLX for Apple silicon

FormatSizeMemory neededRepo
MLX 4bitlmstudio-community/Qwen3.8-27B-MLX-4bit15.0 GBfile size16.0 GB15.0 + 0.50 + 0.50Download page
MLX 8bitlmstudio-community/Qwen3.8-27B-MLX-8bit27.5 GBfile size28.5 GB27.5 + 0.50 + 0.50Download page

Original weights size from the safetensors header, for vLLM, SGLang, Transformers

FormatSizeMemory neededRepo
BF16original weights51.7 GBfile size52.7 GB51.7 + 0.50 + 0.50
hf download Qwen/Qwen3.8-27B --local-dir Qwen3.8-27B
Other conversions (19; sizes not read)

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

Hugging Face Inference Providers, fetched 19 Nov 2026, 13:58 UTC

23 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.

ProviderInOutContextPrecisionTokens/sFirst tokenUptime 24hToolsJSON
Near AI$0.04$1.35256kfp8731.4 s99.9%yesyes
Cerebras$0.99$1.4964kfp164130.4 s99.9%yesno
Mancer 2$0.13$1.50256kfp8690.9 s88.9%yesyes
Tenstorrent$0.15$1.60256kfp81000.7 s100.0%yesyes
DeepInfra$0.15$1.88256kfp8360.7 s99.9%yesyes
Ionstream$0.04$1.95256kfp8950.4 s99.2%yesyes
AkashML$0.23$1.98256kfp8360.9 s99.4%yesyes
Parasail$0.24$2.20256kfp8400.6 s100.0%yesyes
Wafer$0.06$2.20256kfp82390.4 s100.0%yesyes
Modal$0.04$2.30256knot given1090.4 s100.0%yesyes
DekaLLM$0.04$2.50256knot given1170.8 s100.0%yesyes
ModelRun$0.20$2.50256kfp42000.2 s100.0%yesyes
nscale$0.20$2.501Mnot given1170.4 s?yesyes
Reka$0.15$2.50256kfp8740.7 s100.0%yesyes
SiliconFlow$0.42$2.55256kfp8411.0 s100.0%yesno
PrimeIntellect$0.36$2.57256kfp882.9 s100.0%yesyes
CoreWeave$0.40$3.00256kfp8890.5 s99.6%yesyes
Nebius$0.45$3.00977kfp42090.8 s100.0%yesyes
Novita$0.42$3.00977knot given440.8 s100.0%yesno
ovhcloud$0.47$3.19256knot given280.6 s?yesyes
Cloudflare$0.45$3.20256knot given310.7 s100.0%yesyes
Venice$0.45$3.20256kfp8351.1 s100.0%yesyes
featherless-ai???not given???not givennot 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/Qwen3.8-27B:near-ai", "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

from our snapshot of the Hub

Repos that declare this model as their base, most downloaded first. Listing is not a recommendation; unreviewed ones have not been checked.

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.