Qwen3.6 Plus: 1M Token Context, Completely Free

Qwen3.6 Plus Preview dropped on OpenRouter on March 30, 2026, and the headline number is real: 1,000,000 token context window with $0 input and $0 output pricing on the free tier.

The tradeoff: it's a data-for-compute deal. OpenRouter's free tier logs prompts and completions to improve the model — so don't feed it sensitive IP. The model itself is no slouch — hybrid dense/MoE architecture, mandatory chain-of-thought reasoning, full tool use and structured output support. Positioned for agentic coding and complex problem-solving.

For context: paying models charge real money for far less headroom. A 1M-token context lets you stuff entire codebases, long documents, or hours of transcripts into a single prompt — for free, from an open-weights lab.

Why it matters: open-source AI just commoditized long-context inference. The gap between frontier paid models and free alternatives keeps shrinking — fast.