Qwen3 Next 80B A3B Instruct Drops Output Price by 29%

What Changed

Qwen has trimmed API pricing on Qwen3 Next 80B A3B Instruct. Input tokens slip slightly from $0.1 to $0.0975 per million, but the bigger move is on the output side: $1.10 down to $0.78 per million tokens — a -29% reduction overall.

Who Actually Feels This

Output-heavy workloads are the clear winners. If you're running long-form generation, multi-step reasoning chains, or agentic loops where the model produces substantially more tokens than it consumes, that drop from $1.10 to $0.78 compounds fast. At moderate scale — say, 100M output tokens a month — you're looking at $32 back in your budget without changing a single line of code.

For mostly retrieval or classification tasks where output is short, the input cut from $0.1 to $0.0975 is real but marginal.

Context

Qwen3 Next 80B A3B is a mixture-of-experts architecture, activating roughly 3B parameters per forward pass despite the 80B total. That efficiency is part of why pricing can move this aggressively. The -29% cut suggests either improved inference infrastructure on Qwen's end or competitive pressure from other capable MoE models in the same tier.

Check the Qwen3 Next 80B A3B Instruct — live specs & price history page to compare against alternatives and track whether further cuts follow — this model has moved before.

Bottom Line

If output volume drives your costs, reprice your workloads now. This is one of the more substantial single-model output cuts we've logged recently, and it makes the 80B A3B a stronger candidate against pricier dense-model alternatives in the same capability bracket.

Ezra, Scout AI Team