The History of AI

Claude Opus 4.6

Era 5 · Generative AI, reasoning, and agents (2022–2026) › Frontier models

Lineage
Neurons and networks
Kind
Implementation

Who

Anthropic

What happened

Anthropic releases Claude Opus 4.6, the first Opus model with a million-token context window. It brings agent teams, which work in parallel and coordinate on their own.

Why it mattered

Teams of agents that divide a task among themselves. A million tokens of context in an Opus model.

The full story

On February 5, 2026, Anthropic released Claude Opus 4.6. It kept its predecessor's price and added two new features aimed at long tasks.

The first was a million-token context window, in beta and for the first time in an Opus model, which makes it possible to handle much longer texts in one go.

The second was agent teams in Claude Code, Anthropic's coding tool: several agents work in parallel as a team and coordinate autonomously. In one experiment, 16 agents wrote a C compiler in Rust capable of compiling the Linux kernel, at a cost of about $20,000.

According to Anthropic, Opus 4.6 posted the top score on Terminal-Bench 2.0, an agentic coding benchmark, and beat OpenAI's GPT-5.2 by about 144 Elo points on GDPval-AA, which evaluates office work in fields such as finance and law. Claude in PowerPoint also arrived, as a preview.

Two weeks later Google released Gemini 3.1 Pro, and a month later OpenAI released GPT-5.4.

Sources

How to cite

“Claude Opus 4.6.” The History of AI, revised November 13, 2026. https://thehistoryofai.com/en/milestone/claude-opus-4-6

This entry was last reviewed on November 13, 2026.

Open in the interactive timeline