DeepSeek V4-Pro API Prices Surge Up to 12x

🤖 AI Tool Alert — Aug 16, 2026

DeepSeek V4-Pro API Prices Surge Up to 12x

DeepSeek introduced time-based peak/off-peak API pricing effective August 17, alongside the GA release of V4-Pro 0813 (Claude-level agent performance, Responses API support).

Pricing shift:

• Cache-hit input: Off-peak RMB 0.15/M → Peak RMB 0.30/M (12x old flat rate of RMB 0.025)


• Cache-miss input: Off-peak RMB 4.50/M → Peak RMB 9.00/M


• Output: Off-peak RMB 13.50/M → Peak RMB 27.00/M


• Peak hours: 01:00–04:00 & 06:00–10:00 UTC

Why it matters:
1. Your costs now depend on when you run things — early morning UTC hits peak pricing for US devs
2. Off-peak is still competitive (~$1.85/M output), but the "swap and save" margin is gone
3. Batch and agent workloads need a schedule rethink — same playbook as AWS Reserved Instances

💰 Concrete savings example — agent code review pipeline:
A team doing 20+ code reviews/day (~2M output tokens each):

• Peak pricing: ~$3.70/review


• Off-peak pricing: ~$1.85/review


• Fix: Wrap agent calls in a scheduler (GitHub Actions cron: or Temporal) for off-peak UTC


• Savings: ~$1,110/month

Action: Audit your DeepSeek API usage. Add a time check or queue to route non-urgent requests to off-peak. For urgent requests, factor peak pricing into your cost model.

🔗 API Pricing | V4-Pro 0813 Release

#ai #deepseek #apipricing #costoptimization