Key takeaways

• Qwen3.6 27B is a 27-billion-parameter Transformer model from Alibaba, designed for conversational AI, accepting text, image, and video inputs to generate text responses.
• The model demonstrates strong mathematical problem-solving skills across various domains, achieving 97.33% accuracy on the MATH-500 benchmark and 50% Exact Match Accuracy on AIME 2026.
• Capabilities include excellent performance in Algebra, Geometry, Number Theory, and Intermediate Algebra tasks, applying complex mathematical concepts like Simon's Favorite Factoring Trick and properties of permutations.
• Limitations include occasional formatting errors in mathematical outputs, misinterpretations of geometric and combinatorial conditions, and struggles with highly abstract or multi-step symbolic manipulations. The model also frequently failed to provide conclusive answers for complex AIME problems despite detailed steps.
• The model had a 97.33% accuracy on the MATH-500 benchmark, showing strengths in consistent high accuracy across varied math topics. Its lower 50% Exact Match Accuracy on AIME 2026 indicates challenges with intricate combinatorial counting, multi-step geometric transformations, and reaching conclusive results for very complex problems.

Technical specifications

Context length

262k tokens

Max output

—

Architecture

Transformer

Modalities

text+image+video->text

Parameters

~27B

Open weights

Yes

Licence

Apache 2.0

Regions

China

Vendor

Qwen

Provider model key

qwen/qwen3.6-27b