Key takeaways

• Qwen3 Coder Plus is an Alibaba-developed, proprietary text-to-text Mixture of Experts (MoE) model specifically designed for autonomous programming, leveraging tool calling and environment interaction.
• The model's architecture is MoE with 175 billion parameters and a substantial context length of 128,000 tokens, which is particularly beneficial for managing extensive codebases.
• Key capabilities include strong coding proficiency combined with versatile general-purpose abilities, and it demonstrated perfect accuracy on the MATH subset benchmark (100% on 100 problems across various mathematical domains and difficulty levels).
• A significant limitation is its proprietary nature, meaning no open weights are available. Information regarding training methodology, detailed evaluation, or best/worst-case scenarios has not been disclosed, making performance quantification difficult.
• Despite its strength in some areas, the model showed weaknesses in advanced combinatorial problems, geometric reasoning, and complex function analysis, achieving only 40% accuracy on the AIME 2025 benchmark.

Technical specifications

Context length

128k tokens

Max output

—

Architecture

MoE

Modalities

text->text

Parameters

~175B

Open weights

No

Licence

Proprietary

Regions

China

Vendor

Qwen

Provider model key

qwen/qwen3-coder-plus