NEWHow AI models draw the world map · Gecko Tests
Beta
Home/Models/Qwen3 30B A3B Thinking 2507
Alibaba Qwen logo

Qwen3 30B A3B Thinking 2507

by Alibaba Qwen · Released Aug 2025

Open Source
63.5
avg score
Rank #69
Compare
Better than 78% of all models
Context
82K tokens (~41 books)
Input $/1M
$0.20
Output $/1M
$2.40
Type
text
License
Open Source
Benchmarks
6 tested
Data as of
About

Qwen3-30B-A3B-Thinking-2507 is a 30B parameter Mixture-of-Experts reasoning model optimized for complex tasks requiring extended multi-step thinking. The model is designed specifically for “thinking mode,” where internal reasoning traces are separated...

Tested on 6 benchmarks · BenchGecko score 63.5. Top scores: OpenCompass — IFEval (89.7%), OpenCompass — AIME2025 (86.8%), OpenCompass — MMLU-Pro (79.5%).

Capabilities
coding
59.4
#50 globally
math
86.8
#14 globally
knowledge
53.8
#96 globally
language
89.7
#16 globally
Benchmark Scores
Compare All
Tested on 6 benchmarks · Ranked across 4 categories
Score Distribution (all 312 models)
0255075100
▲ You are here
codingCompare coding →
OpenCompass — LiveCodeBenchV6

OpenCompass Live Code Bench v6. Fresh competitive programming problems to evaluate code generation without memorization.

59.4·
mathCompare math →
OpenCompass — AIME2025

OpenCompass evaluation on AIME 2025 problems. Tests mathematical reasoning on fresh competition problems.

86.8·
knowledgeCompare knowledge →
OpenCompass — MMLU-Pro

OpenCompass MMLU-Pro evaluation. Harder knowledge test with more answer choices.

79.5·
OpenCompass — GPQA-Diamond

OpenCompass evaluation of GPQA Diamond. PhD-level science questions from the hardest subset.

70.1·
OpenCompass — HLE

OpenCompass evaluation of Humanitys Last Exam. Expert-level cross-discipline knowledge test.

11.7·
Quick compare:moonshotai logovs Kimi K2.7 CodeMistral AI logovs Mixtral 8x7BOpenAI logovs GPT-5.1-Codex-Mini
Excellent (85+) Good (70-85) Average (50-70) Below (<50)
Model Family · Alibaba Qwen Qwen 3
Apr 2025
39.2
$0.12/M in131Kctx6 benchmarks
Apr 2025
55.2+16.0
$0.46/M in(+0.34)131Kctx10 benchmarks
Jul 2025
44.9-10.3
$0.09/M in(-0.36)262Kctx(+131K)22 benchmarks
Jul 2025
55.3+10.4
$0.23/M in(+0.14)131Kctx(-131K)27 benchmarks
Apr 2025
34.0-21.3
$0.12/M in(-0.11)131Kctx9 benchmarks
Jul 2025
40.1+6.1
$0.05/M in(-0.07)262Kctx(+131K)7 benchmarks
Aug 2025
63.5+23.4
$0.20/M in(+0.15)82Kctx(-180K)6 benchmarks
Apr 2025
47.8-15.7
$0.08/M in(-0.12)131Kctx(+49K)14 benchmarks
Apr 2025
$0.00/M in(-0.08)41Kctx(-90K)
Apr 2025
36.3-11.5
$0.12/M in(+0.12)131Kctx(+90K)12 benchmarks
Jul 2025
$0.07/M in(-0.05)262Kctx(+131K)
Jul 2025
$0.30/M in(+0.23)262Kctx
Jul 2025
$0.00/M in(-0.30)1.0Mctx(+786K)3 benchmarks
Sep 2025
$0.20/M in(+0.20)1.0Mctx(-49K)
Feb 2026
$0.12/M in(-0.08)262Kctx(-738K)12 benchmarks
Sep 2025
$0.65/M in(+0.53)1.0Mctx(+738K)
Sep 2025
49.2+12.9
$0.78/M in(+0.13)262Kctx(-738K)13 benchmarks
Feb 2026
$0.78/M in262Kctx3 benchmarks
Sep 2025
49.2
$0.09/M in(-0.69)262Kctx26 benchmarks
Sep 2025
$0.00/M in(-0.09)262Kctx3 benchmarks
Sep 2025
56.8+7.6
$0.15/M in(+0.15)262Kctx20 benchmarks
Sep 2025
$0.21/M in(+0.06)262Kctx1 benchmark
Sep 2025
$0.40/M in(+0.19)131Kctx(-131K)1 benchmark
Oct 2025
$0.15/M in(-0.25)262Kctx(+131K)
Oct 2025
$0.20/M in(+0.05)262Kctx
Oct 2025
$0.10/M in(-0.10)131Kctx(-131K)
Oct 2025
$0.12/M in(+0.01)262Kctx(+131K)
Oct 2025
$0.18/M in(+0.06)131Kctx(-131K)
See the full Qwen 3 family →
Recently Happened
Qwen3 30B A3B Thinking 2507 pricing increased 290%
Jul 2, 2026
Similar Models
moonshotai logo
Kimi K2.7 Code
moonshotai
63.6$0.61/1M
Mistral AI logo
Mixtral 8x7B
Mistral AI
63.6n/a
OpenAI logo
GPT-5.1-Codex-Mini
OpenAI
63.3$0.25/1M
Links
Info
Alibaba QwenPricing explorerDevelopers · API
Research
Technical Report
Documentation
API DocsPlayground
Community
@Alibaba Qwen
Source Code
GitHubHugging Face
BenchGecko API
qwen3-30b-a3b-thinking-2507
Specifications
  • Typetext
  • Context82K tokens (~41 books)
  • ReleasedAug 2025
  • LicenseOpen Source
  • StatusActive
  • Cost / Message~$0.003
Available On
Alibaba Qwen logoAlibaba Qwen$0.20
Categories
codingmathknowledgelanguage
Learn More
transformeropen weightstokens
Share & Export
Tweet
Related Models
moonshotai logoKimi K2.7 Code
63.6
Mistral AI logoMixtral 8x7B
63.6
OpenAI logoGPT-5.1-Codex-Mini
63.3
OpenAI logoGPT-4 (older v0314)
63.8
Qwen3 30B A3B Thinking 2507 is an open-source text AI model by Alibaba Qwen, released in August 2025. It has an average benchmark score of 63.5. Context window: 82K tokens.

Related Models

Kimi K2.7 Code · moonshotaiMixtral 8x7B · Mistral AIGPT-5.1-Codex-Mini · OpenAIGPT-4 (older v0314) · OpenAIGLM 5.1 · z-ai

Benchmarks

OpenCompass — IFEvalOpenCompass — AIME2025OpenCompass — MMLU-ProOpenCompass — GPQA-DiamondOpenCompass — LiveCodeBenchV6

Related Pages

Alibaba Qwen · ProviderAlibaba Qwen · EconomyAll ModelsCompare ModelsPricingDevelopers · API

Key facts · as of 2026-10-05

  • Qwen3 30B A3B Thinking 2507 by Alibaba Qwen. BenchGecko score 63.5, rank 69 of 312 scored models (normalized average of public benchmark scores).
  • List price $0.20 input · $2.40 output per 1M tokens (as of 2026-10-05).
  • Sold by 1 provider (as of 2026-10-05): Alibaba $0.20 in / $2.40 out.

How to cite · data as of 2026-10-05

Qwen3 30B A3B Thinking 2507 · benchmarks, pricing and providers. BenchGecko, data as of 2026-10-05. https://benchgecko.ai/model/qwen3-30b-a3b-thinking-2507

Credit "Source: BenchGecko" with a link. Prices per provider and Gecko Tests are BenchGecko data (CC BY 4.0); benchmark scores keep their original source, listed in the JSON. JSON · llms.txt · MCP