DeepSeek-V4 Preview Released
DeepSeek-V4 preview is now available for early access. Featuring enhanced multi-modal capabilities and improved reasoning performance.
Loading...
Stay up-to-date with the latest DeepSeek announcements, model releases, and community updates.
DeepSeek-V4 preview is now available for early access. Featuring enhanced multi-modal capabilities and improved reasoning performance.
DeepSeek-V3.2 is officially released with significant improvements in coding, math reasoning, and instruction following.
Experimental version of V3.2 released for community testing and feedback.
DeepSeek V3.1 receives a performance update with better function calling and JSON mode support.
DeepSeek V3.1 introduces improved context handling and faster inference speeds.
Updated R1 model with enhanced chain-of-thought reasoning and reduced hallucination.
V3 model update with optimized performance for coding and mathematical reasoning tasks.
DeepSeek-R1, our most advanced reasoning model, is released with chain-of-thought capabilities for complex problem solving.
Official DeepSeek mobile and desktop applications launched for iOS, Android and Web.
DeepSeek-V3 achieves state-of-the-art performance, rivaling top-tier models at a fraction of the cost.
Minor update to V2.5 with bug fixes and performance improvements.
Lightweight version of the R1 reasoning model for cost-effective deployment.
DeepSeek-V2.5 brings major improvements in multi-turn conversation and code generation.
Disk caching feature launched for the DeepSeek API, reducing latency for repeated queries.
DeepSeek API upgraded with new features including JSON mode, function calling, and prefix caching.