DeepSeek V4 Exits Preview, Adds Peak-Hour Pricing That Doubles API Rates
DeepSeek's V4-Pro and V4-Flash models moved from preview to full production status on July 24, retiring the legacy deepseek-chat and deepseek-reasoner aliases and introducing time-of-day pricing that doubles costs during Beijing business hours.
DeepSeek set July 24, 2026, 15:59 UTC as the cutoff back in April when it first previewed V4, and that deadline landed today: the `deepseek-chat` and `deepseek-reasoner` aliases, which had quietly routed to V4-Flash since the preview, are now fully retired in favor of direct V4 model names.
- V4-Pro: 1.6T total / 49B active parameters, MIT-licensed open weights, scores 80.6% on SWE-bench Verified, within 0.2 points of Claude Opus 4.6
- V4-Flash: 284B total / 13B active parameters, also open-weighted, both support a 1M-token context window
- New peak-hour pricing doubles rates from 9am-12pm and 2pm-6pm Beijing time: V4-Pro output goes from $0.87 to $1.74 per million tokens, V4-Flash from $0.28 to $0.56
- V4-Pro input pricing on a cache miss holds at $0.435 per million tokens off-peak
The peak/off-peak split is the first time a major model provider has tied API pricing to time of day rather than a flat rate, a move that could push latency-tolerant batch workloads toward off-peak hours. Compare V4-Pro's coding scores against other open-weights models in /coding.