Qwen3-235B-A22B-Thinking-2507

Name: Qwen3-235B-A22B-Thinking-2507
Author: Alibaba

Alibaba

Qwen3-235B-A22B-Thinking-2507 is an advanced thinking-mode model on the Mixture-of-Experts architecture with 235B total parameters (22B active). Features 94 layers, 128 experts (8 active), and supports a native context length of 262K. This version is significantly improved in reasoning, achieving leading results among open models with thinking mode in logic, math, science, coding, and academic benchmarks.

Key Specifications

Parameters

235.0B

Context

256.0K

Release Date

July 24, 2025

Average Score

69.2%

API Documentation Repository Model Weights Results Blog

Timeline

Key dates in the model's history

Announcement

July 24, 2025

Last Update

January 22, 2026

Today

September 10, 2026

Technical Specifications

Parameters

235.0B

Training Tokens

Knowledge Cutoff

Family

Capabilities

MultimodalZeroEval

Pricing & Availability

Input (per 1M tokens)

$0.30

Output (per 1M tokens)

$3.00

Max Input Tokens

256.0K

Max Output Tokens

131.1K

Supported Features

Function CallingStructured OutputCode ExecutionWeb SearchBatch InferenceFine-tuning

Benchmark Results

Model performance metrics across various tests and benchmarks

Reasoning

Logical reasoning and analysis

GPQA

Self-reported by the model provider • Self-reported

81.1%

Other Tests

Specialized benchmarks

MMLU-Redux

• Self-reported

94.0%

AIME 2025

• Self-reported

92.0%

IFEval

• Self-reported

88.0%

Arena-Hard v2

GPT-4 evaluated win rates • Self-reported

79.7%

BFCL-v3

Self-reported by the model provider • Self-reported

71.9%

CFEval

Self-reported by the model provider • Self-reported

21.3%

Creative Writing v3

Self-reported by the model provider • Self-reported

86.1%

HMMT25

Self-reported by the model provider • Self-reported

83.9%

Humanity's Last Exam

text-only subset • Self-reported

18.2%

Include

Self-reported by the model provider • Self-reported

81.0%

LiveBench 20241125

Self-reported by the model provider • Self-reported

78.4%

LiveCodeBench v6

25.02-25.05 • Self-reported

74.1%

MMLU-Pro

Self-reported by the model provider • Self-reported

84.4%

MMLU-ProX

Self-reported by the model provider • Self-reported

81.0%

Multi-IF

Self-reported by the model provider • Self-reported

80.6%

OJBench

Self-reported by the model provider • Self-reported

32.5%

PolyMATH

Self-reported by the model provider • Self-reported

60.1%

SuperGPQA

Self-reported by the model provider • Self-reported

64.9%

Tau2 Airline

Self-reported by the model provider • Self-reported

58.0%

Tau2 Retail

Self-reported by the model provider • Self-reported

71.9%

Tau2 Telecom

Self-reported by the model provider • Self-reported

45.6%

TAU-bench Airline

Self-reported by the model provider • Self-reported

46.0%

TAU-bench Retail

Self-reported by the model provider • Self-reported

67.8%

WritingBench

Self-reported by the model provider • Self-reported

88.3%

License & Metadata

License

apache-2.0

Announcement Date

July 24, 2025

Last Updated

January 22, 2026

Similar Models

All Models

Qwen3 235B A22B

Alibaba

235.0B

Best score:0.9 (MMLU)

Released:Apr 2025

Price:$0.20/1M tokens

Qwen3-235B-A22B-Instruct-2507

Alibaba

235.0B

Best score:0.8 (GPQA)

Released:Jul 2025

Price:$0.15/1M tokens

Qwen3.5 122B A10B

Alibaba

122.0B

Best score:0.9 (GPQA)

Released:Mar 2026

Qwen3-Coder 480B A35B Instruct

Alibaba

480.0B

Best score:0.8 (TAU)

Released:Jan 2025

Qwen3 Max Thinking

Alibaba

1.0T

Best score:0.9 (GPQA)

Released:Feb 2026

Price:$1.20/1M tokens

Qwen2.5-Coder 7B Instruct

Alibaba

7.0B

Best score:0.9 (HumanEval)

Released:Sep 2024

Qwen2.5 14B Instruct

Alibaba

14.7B

Best score:0.8 (HumanEval)

Released:Sep 2024

Qwen2 72B Instruct

Alibaba

72.0B

Best score:0.9 (HumanEval)

Released:Jul 2024

Recommendations are based on similarity of characteristics: developer organization, multimodality, parameter size, and benchmark performance. Choose a model to compare or go to the full catalog to browse all available AI models.