Qwen3 30B A3B Thinking 2507 API Prices Rise 54% — What It Costs Now

The Numbers

Qwen's Qwen3 30B A3B Thinking 2507 just got more expensive. Input pricing has moved from $0.13 to $0.20 per million tokens, and output pricing from $1.56 to $2.40 per million tokens — a +54% increase across the board. Check the full Qwen3 30B A3B Thinking 2507 — live specs & price history for the updated rate timeline.

Does It Matter for Your Workload?

The answer depends heavily on your input/output ratio. Thinking-mode models like this one tend to generate verbose reasoning traces, which means output tokens dominate your bill. At $2.40 per million output tokens, a workload that was costing $100/month now costs roughly $154.

For low-volume or experimental use, the delta is negligible. For production pipelines running thousands of requests daily — particularly anything doing multi-step reasoning, code generation, or chain-of-thought tasks — this is worth a budget line review.

Practical Takeaways

• Recalculate at scale. If you benchmarked costs at the old rates, rerun your estimates. The +54% jump is material at volume.
• Output is the lever. Trimming system prompts and capping max tokens will have more impact than anything else right now.
• Compare alternatives. Our models index lists comparable mid-size MoE and reasoning models with current pricing — worth a side-by-side before committing.

This model remains competitively positioned in the thinking/reasoning category, but the price floor has shifted. Factor it into any cost-per-task analysis before your next infrastructure decision.

Ezra, Scout AI Team