Qwen3-235B-A22B-Thinking (Jul 2025)

AI model Released Jul 25, 2025

Open weights: Yes^*

Capability & preference

General capability · ECI

A cross-benchmark capability estimate. Index points, not a percentage; intervals show uncertainty.

Qwen3-235B-A22B-Thinking (Jul 2025) 143.2^* Qwen3-235B-A22B-Thinking (Jul 2025) · 95% interval 141.3–145.0

Model performance

Radar task score: four tests, equal weight. An experimental summary of these tasks, not a universal intelligence rating.

Qwen3-235B-A22B-Thinking (Jul 2025)

Tested configuration Radar task score

Unrated

3/4 core tests · not enough evidence

4 benchmark results for this configuration

Scientific reasoning

GPQA Diamond · 25%

Expert-written science questions; strict multiple-choice scoring. Chance accuracy is 25%.

Qwen3-235B-A22B-Thinking (Jul 2025) 80.1^*% Approx. 95% interval: 75.0–85.1% · 2025-12-11 · Older result

Novel problem solving

Mystery Game Puzzles · 25%

Best-next-move puzzles in an undisclosed game variant. A narrow reasoning test, not general intelligence.

Qwen3-235B-A22B-Thinking (Jul 2025) 9.0^*% Approx. 95% interval: 3.4–14.6% · 2026-08-27

Advanced mathematics

FrontierMath Tiers 1-3 · 25%

Private v2 mathematics problems with Python tools. Earlier versions and Tier 4 are excluded.

Qwen3-235B-A22B-Thinking (Jul 2025) Not available

Factual answers

SimpleQA Verified · 25%

Correct factual answers under a guessing prompt, graded by a model. Not a hallucination or abstention score.

Qwen3-235B-A22B-Thinking (Jul 2025) 40.4^*% Approx. 95% interval: 37.4–43.5% · 2026-08-27

Competition mathematics

OTIS Mock AIME 2024-2025 · Separate result

Competition problems with model-extracted answers checked against an exact key. Earlier grading procedures are excluded.

Qwen3-235B-A22B-Thinking (Jul 2025) 86.7^*% Approx. 95% interval: 76.6–96.7% · 2025-12-10 · Older result

Bars show reported task scores. Whiskers are approximate 95% intervals where standard errors are available. Different reasoning budgets are not equal-cost tests. Missing results are never zero.

Run it locally

Inference · memory estimate

Qwen3-235B-A22B-Thinking (Jul 2025)

Not enough verified architecture information for a reliable VRAM estimate.

Capabilities

Open weights: Yes^*

System configuration

Detailed specifications

Specifications · 4 fields

Access model: Open weights (unrestricted)^*
Open weights: Yes^*
Organization: Alibaba^*
Organization country: China^*

* Provisional data