DeepSeek on Thursday released the official version of its flagship DeepSeek-V4-Pro model while sharply raising prices for its application programming interface, with some rates increasing by as much as 1,100 percent.
The Chinese AI startup introduced peak and off-peak pricing tiers that take effect August 16. During peak hours from 9 a.m. to noon and 2 p.m. to 6 p.m. Beijing time, API prices for DeepSeek-V4-Pro now stand at 0.3 yuan (about $0.04) per million tokens for cached input, 9 yuan for uncached input, and 27 yuan for output. Off-peak prices are set at half those rates.
The previous flat-rate pricing was just 0.025 yuan for cached input, 3 yuan for uncached input, and 6 yuan for output, according to Caixin Global. The dramatic increase signals a shift away from DeepSeek’s earlier strategy of undercutting competitors on price to build market share.
Stronger Agent Capabilities
The official V4-Pro release emphasizes stronger agent capabilities, with DeepSeek claiming the model outperforms xAI’s GLM-5.2 in several agent benchmarks and beats Anthropic’s Claude Opus 4.8 in selected tests. However, it trails Kimi K3 and Claude Fable 5 in most evaluations.
On the third-party evaluation platform Artificial Analysis, DeepSeek-V4-Pro scored 53, tying GLM-5.2 and ranking behind models including Opus 5, Fable 5, GPT-5.6 Sol, Grok 4.6, Kimi K3, and Qwen3.8-Max. The model also adds native support for OpenAI’s Responses API format.
A Pivotal Pricing Shift
The price hike is notable because DeepSeek had previously positioned itself as the budget option in the frontier model market, undercutting Western rivals by a wide margin. The new pricing still remains below most Western competitors, but the 1,100 percent increase for certain tiers marks a fundamental change in the company’s approach.
The move comes amid broader turbulence in AI pricing as companies grapple with the costs of running increasingly capable models. The introduction of peak and off-peak tiers mirrors electricity pricing models, reflecting the infrastructure costs of serving high-demand AI workloads.
DeepSeek announced the changes alongside its V4-Pro launch on August 14, with the new pricing tiers going into effect two days later. The company did not immediately respond to requests for comment on the rationale behind the sharp increases.
Sources: Caixin Global; Artificial Analysis
discussion