DeepSeek released DeepSeek-V4-Flash (0731) and updated API pricing with peak/off-peak rates for its V4 models

Decision brief

What changed

DeepSeek moved DeepSeek-V4-Flash-0731 from preview to broad availability and published lower list API prices for it at $0.14 per 1 million input tokens and $0.28 per 1 million output tokens. Separately, DeepSeek announced peak/off-peak API pricing for its V4 models effective Aug. 16, 2026, with off-peak at half the regular rate and reported peak-hour output pricing for V4-Flash rising to $1.32 per 1 million tokens.

Why it matters

For leaders buying or routing AI workloads, this changes both model-selection economics and run-time scheduling economics: coverage says V4-Flash may match or beat V4-Pro on some coding and agentic tasks while starting at roughly one-third of V4-Pro’s listed token price. The new time-of-day pricing means unit costs can vary sharply by usage window, so teams with flexible batch or agent workloads may be able to lower spend by shifting demand off-peak, while latency-sensitive or always-on use cases could face materially higher peak costs.

Affected roles

CFO COO CTO

Evidence

The release and base pricing were reported consistently by The New Stack and The Batch, both citing DeepSeek’s published benchmarks and pricing; The Batch also referenced Artificial Analysis results showing V4-Flash ahead of V4-Pro on its index. The pricing-schedule change was separately reported by Trending Topics EU and Fortune, with Fortune specifying the peak-hour V4-Flash output-token increase and the Aug. 16 effective date.

What remains uncertain

The coverage does not establish how V4-Flash will perform across your specific workloads, nor whether its apparent advantage over V4-Pro in reported tests will hold outside the cited benchmarks and examples. It is also unclear from the coverage how broad the peak windows are across regions, what total blended cost will look like under real traffic patterns, and whether developers can reliably shift enough volume to off-peak periods to realize savings.

Monitor next

Watch for DeepSeek’s final production pricing documentation and customer-reported usage data after Aug. 16, especially any published peak-window definitions and real-world cost/performance comparisons between V4-Flash and V4-Pro.

Analytical support, not advice — assumptions and open questions stated above.

Source coverage