Qwen
Qwen
Language model · China ·Apr 2025

Qwen3-4B

Language Open (Restricted)
4B
Parameters
46
Benchmarks

About Qwen3-4B

Qwen3-4B is an AI model developed by Qwen, in the language category, released in 2025, made available as an open-weights model with 4B params.

On this page you'll find Qwen3-4B's full specifications, including 46 benchmark results. Review provider pricing and benchmark scores below, or compare Qwen3-4B head-to-head with other language models.

Links

Run Qwen3-4B locally — what to buy

Share this build ↗

Sized to what Qwen3-4B actually needs (~8k context at Q4), not to the biggest GPU: Good = cheapest that runs it, Better = best value, Best = most headroom. Speed is a hardware estimate; anything past ~120 tok/s is shown as instant.

Software support

✓ measured · · compatible · — not supported. Informational only — speed is hardware-based.

Community

Would you run Qwen3-4B again?
No votes yet — be the first.
Did it run? — community reports by hardware

No reports yet. Own this model on some hardware? Be the first to confirm it runs.

Discussion

No comments yet. Start the discussion.

Qwen3-4B benchmark scores
Benchmark Category Score Variant
Codeforces Elo code 1,671 thinking-rating
Codeforces Elo code 842 non-thinking-rating
LiveCodeBench v5 code 45.5% thinking-pass@1
LiveCodeBench v5 code 21.3% non-thinking-pass@1
AlignBench v1.1 general 8.3 thinking-v1-1
AlignBench v1.1 general 8.1 non-thinking-v1-1
ArenaHard general 76.6% thinking
ArenaHard general 66.2% non-thinking
C-Eval general 77.5% thinking
C-Eval general 72.2% non-thinking
Creative Writing v3 general 61.1 thinking
Creative Writing v3 general 53.6 non-thinking
IFEval general 81.9% thinking-strict-prompt
IFEval general 81.2% non-thinking-strict-prompt
LiveBench general 63.6 thinking-2024-11-25
LiveBench general 48.4 non-thinking-2024-11-25
MMLU-Redux general 83.7% thinking
MMLU-Redux general 77.3% non-thinking
MMMLU general 69.8% thinking-14lang
MMMLU general 61.7% non-thinking-14lang
Multi-IF general 66.3 thinking
Multi-IF general 61.3 non-thinking
WritingBench general 7.35 thinking
WritingBench general 6.85 non-thinking
INCLUDE knowledge 61.8 thinking
INCLUDE knowledge 53.8 non-thinking
AIME 2024 math 73.8% thinking
AIME 2024 math 25% non-thinking
AIME 2025 math 65.6% thinking
AIME 2025 math 19.1% non-thinking
MATH-500 math 97% thinking
MATH-500 math 84.8% non-thinking
MT-AIME 2024 math 60.7 thinking-mt-2024
MT-AIME 2024 math 13.9 non-thinking-mt-2024
PolyMath-en math 40 thinking
PolyMath-en math 16.6 non-thinking
AutoLogi reasoning 87.9 thinking
AutoLogi reasoning 76.3 non-thinking
GPQA Diamond reasoning 55.9% thinking
GPQA Diamond reasoning 41.7% non-thinking
MLogiQA reasoning 65.9 thinking
MLogiQA reasoning 49.9 non-thinking
ZebraLogic reasoning 81% thinking
ZebraLogic reasoning 35.2% non-thinking
BFCL v3 tool-use 65.9% thinking
BFCL v3 tool-use 57.6% non-thinking

More models like Qwen3-4B

Frequently asked questions

What is Qwen3-4B?

Qwen3-4B is an AI model developed by Qwen, in the language category, released in 2025. It is tracked on GenAIList with its specifications, benchmark scores and provider pricing.

Who created Qwen3-4B?

Qwen3-4B was developed by Qwen and released in 2025.

Is Qwen3-4B open source or proprietary?

Qwen3-4B ships as an open-weights model: you can download and self-host the weights, though the licence may place some restrictions on use.

How much does Qwen3-4B cost?

Pricing for Qwen3-4B depends on the provider. See the providers table on this page for the latest API rates.

How does Qwen3-4B perform on benchmarks?

Qwen3-4B is benchmarked across 46 evaluations on GenAIList, including Codeforces Elo (1,671). See the full benchmark table below and compare it with other models.

No videos yet. Know a good one? Add it below.

Know a great video about Qwen3-4B?