Currently listed through these providers:
Model details
DeepSeek V4 Pro
DeepSeek V4 Pro is a 1.6-trillion parameter Mixture-of-Experts model released in April 2026 under the MIT license, positioned as a flagship entry in DeepSeek's reasoning-oriented lineup. Its design emphasizes advanced reasoning and complex software engineering, drawing on DeepSeek's tradition of building large MoE architectures that route computation efficiently while still supporting very long contexts for production workloads. As a model in the deepseek-thinking family, it is aimed at developers and teams that need sustained chain-of-thought behavior, structured interaction patterns, and the ability to plug into tool calling and temperature-controlled workflows rather than short conversational exchanges.
Because the weights are openly available, organizations can self-host DeepSeek V4 Pro on their own infrastructure, tune sampling and decoders to fit downstream pipelines, and integrate structured output for tasks such as code generation, multi-step planning, and long document analysis. Its scale makes it best suited for demanding reasoning, refactoring, and agent-style coding tasks where larger context budgets and careful tool use pay off, while teams with lighter workloads can rely on smaller members of the same family. For current best results, practitioners should cross-check the model's published behavior against the latest deployment documentation for the runtime they choose.
Quick Info
Powered by- Provider
- CrossModel
- Model key
- deepseek/deepseek-v4-pro
- Release date
- Apr 24, 2026
- Last updated
- Apr 24, 2026
- Knowledge cutoff
- 2025-05
- Input modalities
- Output modalities
- Capabilities
Cost
- Input token cost
- $1.215
- Output token cost
- $3.645
Limits
- Output tokens
- 384,000 tokens
- Context window
- 1,000,000 tokens
Latest news about DeepSeek V4 Pro
Videos about DeepSeek V4 Pro
More models around DeepSeek V4 Pro
This exact model name is also listed by 51 other providers.