DeepSeek cuts V4 Pro pricing by 75% to $0.003625 per million input tokens
DeepSeek has permanently reduced pricing for its V4 Pro model by 75%, bringing input token costs down to $0.003625 per million tokens from $0.0145. The move makes permanent a promotional discount that was set to expire May 31, 2026.
DeepSeek cuts V4 Pro pricing by 75% to $0.003625 per million input tokens
DeepSeek has permanently reduced pricing for its flagship V4 Pro model by 75%, according to an update on the company's website. Input tokens now cost $0.003625 per million, down from $0.0145, while output tokens dropped to $0.87 per million from $3.48.
The price reduction makes permanent a promotional discount that was originally scheduled to end on May 31, 2026. DeepSeek released the V4 Pro and V4 Flash models in April 2026, claiming they would usher in an "era of cost-effective 1M context length."
Pricing comparison
The new pricing structure positions DeepSeek V4 Pro significantly below competing models:
- DeepSeek V4 Pro: $0.003625 input / $0.87 output per 1M tokens
- Previous DeepSeek V4 Pro pricing: $0.0145 input / $3.48 output per 1M tokens
The company describes its positioning as the "cost-effective" choice for AI agents, a strategy that could deliver substantial savings for enterprise accounts and power users processing millions of tokens daily.
Market context
DeepSeek's aggressive pricing comes amid increasing competition in the large language model market. The Chinese startup is now positioned as a lower-cost alternative to OpenAI's GPT-5 and Google's recently released Gemini 3.5 Flash, though specific pricing comparisons for those models were not provided.
The pricing strategy follows previous tensions with competitors. Anthropic has accused DeepSeek of "distillation attacks" — a practice where one company's model improperly learns from another's more capable system. The permanent price cuts may intensify these competitive dynamics.
What this means
DeepSeek's permanent 75% price reduction represents a significant escalation in AI model pricing competition. For enterprise users running high-volume workloads, the cost difference could be substantial — a workload requiring 1 billion tokens per day would now cost approximately $3,625 for input tokens instead of $14,500. However, buyers should evaluate whether DeepSeek's performance matches their requirements, as raw pricing doesn't account for differences in model capabilities, accuracy, or output quality. The move also raises questions about the sustainability of such pricing and whether competitors will respond with their own cuts.
Related Articles
Cline v4.0.2 Adds DeepSeek Reasoning Effort Controls, Including 'xhigh' Setting
Cline, the autonomous AI coding assistant, released v4.0.2 with support for reasoning effort controls on DeepSeek thinking models, including the new 'xhigh' setting. The update also improves the ClinePass provider experience with clearer reasoning controls and model selection.
DeepSeek Releases V4 Models: 1M Context Window, 90% Less KV Cache Than V3
DeepSeek has released two new MoE models: DeepSeek-V4-Pro with 1.6T parameters (49B activated) and DeepSeek-V4-Flash with 284B parameters (13B activated). Both models support a one million token context window and use a hybrid attention architecture that requires only 27% of single-token inference FLOPs and 10% of KV cache compared to DeepSeek-V3.2.
DeepSeek Releases V4-Pro with 1.6T Parameters, 1M Token Context at 27% Inference Cost of V3
DeepSeek has released two Mixture-of-Experts models: V4-Pro with 1.6 trillion parameters (49B activated) and V4-Flash with 284B parameters (13B activated), both supporting 1 million token context windows. V4-Pro requires only 27% of inference FLOPs and 10% of KV cache compared to V3.2 at 1M token context, trained on over 32 trillion tokens.
DeepSeek-V4-Fable: Offensive Security Model Trained on 80,000 CTF Trajectories Achieves 58.7% Solve Rate
Chunjiang Intelligence has released DeepSeek-V4-Fable, an autonomous agent model designed for offensive security research and CTF challenges. The model, distilled from Claude-5-Fable and built on DeepSeek-V4-Flash, was trained on 80,000 verified CTF trajectories and achieves a 58.7% solve rate across held-out security challenges.
Comments
Loading...