Qwen: Qwen3 Next 80B A3B Thinking
🧠 AI Modelqwen
Reasoning-first chat model with 80B params (3B active) and 262k context, outputting structured thinking traces.
Qwen3-Next-80B-A3B-Thinking is a reasoning-first model from the Qwen3-Next line, built for complex multi-step reasoning. It outputs structured thinking traces by default, making it ideal for tasks requiring deep logical analysis, such as mathematical proofs, code synthesis and debugging, and agentic reasoning. The model uses a Mixture-of-Experts (MoE) architecture with 80 billion total parameters, but only 3 billion are activated per token, offering a strong performance-to-cost ratio. It supports a 262,144-token context window and features like frequency penalty, logit bias, min_p, repetition penalty, and configurable reasoning. Pricing is $0.10 per million input tokens and $0.78 per million output tokens. The model excels in benchmarks for reasoning and code tasks.
💡Highlights
- ├─80B total, only 3B activated per token
- ├─262k tokens context length
- └─Structured thinking traces by default
🎯For
- ├─AI researchers
- ├─Developers
- └─Reasoning application builders