Anthropic introduced hybrid reasoning in Claude 3.7 Sonnet (unifying fast responses and controllable extended-thinking budgets in one model) alongside Claude Code, raising the bar on agentic software engineering benchmarks such as SWE-bench.
Anthropic
Also known as: Claude
Plain English
The AI safety and research lab behind the Claude family of models, known for strong software engineering performance, long-context reasoning, mechanistic interpretability research, and creating the Model Context Protocol (MCP).
Technical Overview
Anthropic trains the Claude model series using Constitutional AI and scalable Post-Training (SFT, RLHF, DPO & GRPO) with hybrid extended thinking, long context windows, and prompt prefix caching. Beyond models, Anthropic shaped the agent developer stack by open-sourcing MCP and pioneering agentic coding workflows.
Chronological Entries (2)
Anthropic open-sourced the Model Context Protocol (MCP), proposing a universal client-server interface for connecting LLMs to external tools, databases, and repositories—rapidly adopted across IDEs, LangGraph & Agent Orchestration, OpenAI, and Google DeepMind.
Direct Relations (3)
- EnablesModel Context Protocol (MCP)
- Alternative toOpenAI
- Alternative toGoogle DeepMind