Qwen3 235B A22B Thinking 2507 API Prices Double: Input Hits $0.30, Output $3.00

What changed

Qwen3 235B A22B Thinking 2507 just received a +101% price increase across both input and output. Input tokens move from $0.149 → $0.30 per 1M tokens; output tokens jump from $1.50 → $3.00 per 1M tokens. Both figures have effectively doubled overnight.

Does this matter for your workload?

For low-volume experimentation, the absolute delta is still modest — an extra $0.151 per million input tokens won't break a side project. But at scale the math shifts fast. A pipeline burning 100M output tokens per month goes from $150 to $300 in output costs alone, before touching input. Teams running reasoning-heavy workflows (chain-of-thought, multi-step agents) where output tokens dominate should audit their monthly token spend before the new rate hits their next invoice.

Context worth noting

This model sits in the "thinking" tier, meaning it generates extended reasoning traces — output token counts tend to run higher than standard completions. The $3.00 output rate now places it closer to premium reasoning models from other providers, narrowing the value gap that previously made it a compelling budget pick for MoE-scale reasoning.

What to do next

Check your current token split (input vs. output) and recalculate monthly cost at the new rates. If output tokens dominate and cost is a constraint, it's worth benchmarking against alternatives on our site. Full pricing details and historical rate changes are tracked at Qwen3 235B A22B Thinking 2507 — live specs & price history.

Ezra, Scout AI Team