Alibaba logo

Qwen3-Coder 480B A35B Instruct

Alibaba

Qwen3-Coder-480B-A35B-Instruct is the most agentic code model from Qwen with 480B total parameters (35B active) on the MoE architecture. Achieves significant performance among open models in agentic coding, browser use, and core programming tasks, with results comparable to Claude Sonnet. Native context length of 256K tokens (expandable to 1M with Yarn), optimized for repository-level code understanding.

Key Specifications

Parameters
480.0B
Context
-
Release Date
January 30, 2025
Average Score
54.3%

Timeline

Key dates in the model's history
Announcement
January 30, 2025
Last Update
January 22, 2026
Today
September 10, 2026

Technical Specifications

Parameters
480.0B
Training Tokens
-
Knowledge Cutoff
-
Family
-
Capabilities
MultimodalZeroEval

Benchmark Results

Model performance metrics across various tests and benchmarks

Programming

Programming skills tests
SWE-Bench Verified
With OpenHands, 500Self-reported
70.0%

Other Tests

Specialized benchmarks
TAU-bench Retail
Benchmark agentic tool useSelf-reported
77.5%
BFCL-v3
Benchmark agentic tool useSelf-reported
69.0%
Aider-Polyglot
Benchmark agentic programmingSelf-reported
62.0%
TAU-bench Airline
Benchmark agentic tool useSelf-reported
60.0%
MM-Mind2Web
Agentic Browser Use benchmarkSelf-reported
55.8%
Multi-SWE-Bench
mini variantSelf-reported
25.8%
Spider
Spider2 variantSelf-reported
31.1%
SWE-bench Multilingual
Agentic Coding benchmarkSelf-reported
54.7%
Terminal-Bench 2.0
Agentic Coding benchmarkSelf-reported
37.5%

License & Metadata

License
apache-2.0
Announcement Date
January 30, 2025
Last Updated
January 22, 2026

Similar Models

All Models

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.