Qwen3-Coder-Next: specs, pricing and how it performs in a council

Qwen3-Coder-Next is Alibaba (Qwen)'s model available on Council AI at $0.12 per million input tokens and $0.80 per million output, with a 262,144-token context window. Open-weight Qwen coding model (Feb 2026) optimized for coding agents. Sparse MoE, 80B total / 3B active. 256K context. Replaces Qwen3-Coder, whose qwen3-coder-plus slug OpenRouter removed. On Council you do not run it alone: the same prompt goes to up to 10 models from different labs in parallel, and a moderator synthesizes one answer with a consensus score showing where the labs agreed.

Qwen3-Coder-Next specifications

What Qwen3-Coder-Next is good at

Open-weight Qwen coding model (Feb 2026) optimized for coding agents. Sparse MoE, 80B total / 3B active. 256K context. Replaces Qwen3-Coder, whose qwen3-coder-plus slug OpenRouter removed.

Council surfaces it for: coding, agents, tools.

It has no native web search. Council can still retrieve context for it from your own document library on Ultra.

Qwen3-Coder-Next pricing on Council AI

Council does not resell tokens. Every paid plan carries a monthly budget and each model draws against it at cost: $0.12 in / $0.80 out per million tokens for Qwen3-Coder-Next. Plans start at $19.99/mo (Plus), $39.99/mo (Pro) and $99.99/mo (Ultra, which adds media review, the personal RAG library and MCP server).

If budget is the binding constraint, GPT-6 Luna is the cheaper seat in the lineup at $0.10/$0.50.

Using Qwen3-Coder-Next in a council

A single model — this one included — can be confidently wrong. Council runs Qwen3-Coder-Next alongside models from other labs on the same prompt, then scores how much they agreed. High consensus is a signal you can act on the answer; low consensus surfaces the specific point they split on, which is usually the part worth checking yourself.

Other Alibaba (Qwen) models in the lineup: Qwen3.8-Max, Qwen3.8 Flash.

Frequently asked questions