Qwen: Qwen3 Coder 480B A35B Instruct
Model Type
Proprietary Model
API access only
Recommended Use Cases
Try Qwen3 Coder 480B A35B Instruct
Qwen3 Coder 480B A35B Instruct is Alibaba's flagship open-weight coding model with 480B total parameters and 35B active, achieving state-of-the-art results on agentic coding benchmarks comparable to Claude Sonnet 4.
Our most agentic code model to date. Qwen3-Coder-480B-A35B-Instruct sets new state-of-the-art results among open models on Agentic Coding, Agentic Browser-Use, and Agentic Tool-Use, comparable to Claude Sonnet 4.
- Qwen Team
Overview
Qwen3-Coder-480B-A35B-Instruct is the largest and most capable model in the Qwen3-Coder series. It uses a Mixture-of-Experts architecture with 160 experts (8 activated per token), trained on 7.5T tokens (70% code) with long-horizon reinforcement learning on real-world software engineering tasks.
Key Features
- Flagship capability: State-of-the-art open-source coding model
- 480B MoE: 35B active parameters per forward pass
- 256K native context: Extendable to 1M with YaRN
- Agentic training: 20,000 parallel environments for Agent RL
- SWE-Bench SOTA: Best open-source performance without test-time scaling
Technical Specifications
| Specification | Value |
|---|---|
| Total Parameters | 480B |
| Active Parameters | 35B |
| Experts | 160 (8 activated per token) |
| Architecture | MoE transformer |
| Context Length | 256K tokens (1M with YaRN) |
| Training Data | 7.5T tokens (70% code) |
| Release Date | July 2025 |
| License | Apache 2.0 |
When to Use Qwen3-Coder
Choose Qwen3-Coder-480B when you need:
- Maximum coding capability with open weights
- Complex multi-file refactoring
- Autonomous software engineering tasks
- Self-hosted agentic coding infrastructure
Consider alternatives when you need:
- Efficient local deployment → Qwen3-Coder-Next (80B-A3B)
- API access without self-hosting → Qwen3-Coder-Plus
- Fast, cost-effective coding → Qwen3-Coder-Flash
Availability
- Open Weights: Hugging Face (Qwen/Qwen3-Coder-480B-A35B-Instruct)
- API: Together AI, OpenRouter
- Local: vLLM, SGLang (requires multi-GPU setup)
Role in Series
Qwen3-Coder models by capability:
- Qwen3-Coder-Flash: API, fast and cheap
- Qwen3-Coder-Plus: API, maximum capability
- Qwen3-Coder-30B-A3B: Open-weight, efficient
- Qwen3-Coder-Next: Open-weight, novel architecture
- Qwen3-Coder-480B-A35B: Open-weight flagship (this model)