Breaking

Anthropic Releases Claude Opus 5.5 With Lower Pricing and Agent-Focused Guidance

September 22, 2026 at 16:16 EDT

Foundation Models
AI Agents
Software Dev & Coding

Anthropic released Claude Opus 5.5 on September 22, positioning it as the first model in the Claude 5.5 family for long-running agentic coding and knowledge work. The company says the model approaches Claude Fable 5.1 capability on many tasks while costing 40% less to run than Opus 5, and its developer guidance now emphasizes giving the model a complete task, a clear definition of done and explicit check-in points before letting it work.

Sept. 22, 2026 · Anthropic

Claude Opus 5.5 shifts the agent-cost curve

The new Opus model is aimed at long-running coding and knowledge-work agents, with lower token prices, faster output and guidance to hand over complete tasks with clear finish lines.

40%

lower typical workload cost than Opus 5

$4/$20

per MTok input and output

>30%

faster output generation than Opus 5

token context window

Typical workload cost, indexed

100%Opus 5

60%Opus 5.5

A 40% reduction means a typical Opus 5.5 workload is presented as about three cost blocks where Opus 5 used five.

How Anthropic wants users to run it

Whole task

Give the full objective and context.

→

Done state

Define success and check-ins.

→

Measure

Use real task runs and usage data.

Upside

Lower pricing, faster output, clearer writing and larger practical room for long agent sessions.

Watch points

Always-on thinking, effort settings, cache resets and API breaking changes can change real cost.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.