New DeepSeek-V3.1 marks the dawn of the agent era!
DeepSeek-V3.1 Release - The First Step Toward the Agent Era
Original: DeepSeek-V3.1 Release | DeepSeek API Docs
Importance: 新機能が多くのユーザーに影響を与えるため
Summary
DeepSeek-V3.1 has been released, marking the first step toward the agent era. This version features hybrid inference (two modes: Think and Non-Think), faster answers, and enhanced tool usage. It supports a context length of 128K and improves the API, allowing for more efficient handling of complex search tasks and multi-step agent tasks.
Key Points
- Introduction of hybrid inference feature
- Reduced response time in Think mode
- Enhanced tool usage capabilities
- Support for 128K context length
- Improved API functionality
View developer notes (APIs, breaking changes, migration)
DeepSeek-V3.1 introduces hybrid inference with two modes: Think and Non-Think. The Think mode provides faster responses compared to DeepSeek-R1-0528. Enhanced agent skills improve tool usage and multi-step tasks. The API supports strict function calling in beta, with notable improvements in multi-step reasoning and performance on SWE and Terminal-Bench.
Source: https://api-docs.deepseek.com/news/news250821
Outlet: DeepSeek Release Notes
This article is an AI-generated summary (OpenAI GPT-4o-mini) of publicly available information from Anthropic, OpenAI, Google, Meta, Mistral, DeepSeek, Sakana, and other vendors. The original source URL is always provided in accordance with fair-use citation requirements. Summaries are AI-generated and may contain mistranslations or misinterpretations. Always verify details with the original source.