Qwen3.5 397B A17B Drops Input and Output Prices by 35%

What Changed

Qwen has quietly repriced Qwen3.5 397B A17B. Input tokens dropped from $0.50 to $0.39 per 1M, and output tokens fell from $3.60 to $2.34 per 1M — a flat -35% across both dimensions.

Why It Matters

Output tokens are almost always the cost driver in real applications, so the cut from $3.60 to $2.34 is the number worth watching. For a workload generating 10M output tokens a month, that's roughly $126 saved — before you even count the input savings. At scale, that compounds fast.

This is a large mixture-of-experts model (397B total parameters, 17B active), which means it punches closer to dense-model quality while keeping inference costs manageable. The price cut makes it more competitive against similarly sized open-weight alternatives in the sub-$3 output tier.

Who Should Care

• High-output pipelines (summarization, code generation, long-form drafting): the output price drop is the headline number.
• Cost-sensitive teams already using Qwen3.5 397B A17B: no action required; the lower rate applies automatically via API.
• Evaluators shopping the market: this price cut shifts the value comparison. Check current per-token rates and benchmark comparisons on the Qwen3.5 397B A17B — live specs & price history page before finalising a model choice.

Bottom Line

A 35% price cut on both input and output is a real reduction, not a rounding-error adjustment. If Qwen3.5 397B A17B was close but slightly over budget before, it's worth re-running your cost model now.

Ezra, Scout AI Team