AI Briefing
KO

DeepSeek-R1-0528 Released

Key point

DeepSeek-R1-0528 has been released with improved performance, front-end capabilities, and hallucinations.

1 / 2

Details

DeepSeek-R1-0528 has been unveiled. This update highlights improved benchmark performance, enhanced front-end capabilities, and reduced hallucinations, and it also supports JSON output and function calling.

The key changes are as follows.

  • Improved benchmark performance
  • Enhanced front-end capabilities
  • Reduced hallucinations
  • Support for JSON output and function calling

Users can try it immediately at chat.deepseek.com, and the way the API is used remains unchanged. Related documentation can be found in DeepSeek's thinking mode guide, and the open-source weights are available on Hugging Face.

llmbenchmarkdeepseekfunction-callingjsondeepseek-r1open-source-weights
Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.

Related

DeepSeekNew Open-Source Model DeepSeek-V2.5 Combines General Chat and Coding CapabilitiesRedditOpen-Weight CLM Released as Jev Alternative, Agent Decision-Making 4x FasterAnthropic NewsAnthropic Releases Claude Opus 5.5 with 40% Cost Reduction and Record Alignment ScoresHacker NewsStep 5 Preview Released: 600B Parameter MoE Model Highlights Agentic Tasks and Financial Analysis PerformanceRedditMica v0.1 4B Released: Open-Source Decision Model for Agent Loops Trained for Under $30Hacker NewsQwen3.5-based Small Decision Model 'Kev' Released with TypeSafe API Compatibility and Performance Metrics