DeepSeek unveils DeepSeek-V4.1 Flash topping V4 Pro
DeepSeek announced the scheduled release of its DeepSeek-V4.1 Flash model around September 10, 2026, stating it comprehensively surpasses V4 Pro across performance benchmarks, speed, and cost. Prior to the subsequent release of V4.1 Pro, DeepSeek will reroute all Pro API requests to V4.1 Flash at Flash-tier rates, setting off-peak pricing at $0.003 per million tokens for cache hits and $0.60 for output.
DeepSeek continues its aggressive disruption of frontier model economics, turning conventional tier pricing upside down by delivering flagship performance in a lightweight package.
- –Making a lightweight Flash model outperform the previous flagship Pro tier highlights accelerated efficiency gains from architectural and post-training optimizations.
- –Automatically rerouting Pro requests to V4.1 Flash while lowering the billing rate creates an exceptionally customer-friendly migration path that exerts heavy pricing pressure on competing frontier labs.
- –Enforcing peak vs. off-peak dynamic pricing reflects real-world inference capacity management, setting a utility-style billing precedent for high-demand AI providers.
DISCOVERED
59m ago
2026-09-09
PUBLISHED
6h ago
2026-09-09
RELEVANCE
AUTHOR
nickweb