Claude Opus 5.5

Anthropic · Released 22 Sep 2026 · claude-opus-5-5

Compare this model

ProprietaryCodingAgentsReasoningLong context

$4.00

Input price / 1M tokens

#50 of 61, cheapest first

$20.00

Output price / 1M tokens

#50 of 61, cheapest first

Context window

#29 of 71

Output tokens / second

#31 of 61

66.4%

Terminal-Bench 4.0

#2 of 10

54.4%

FrontierCode 1.1

#1 of 6

40%

AutomationBench

#2 of 8

67.7%

Humanity’s Last Exam (with tools)

#1 of 16

Summary

Claude Opus 5.5 is a proprietary model from Anthropic, released on 22 Sep 2026.

At $4.00 input and $20.00 output per million tokens, it is #50 of 61 on input price, cheapest first.

Its 1M-token context window ranks #29 of 71.

Measured output speed is 93 tokens per second, #31 of 61.

Anthropic's recommended starting model for most workloads, built for long-running agentic coding and knowledge work. Anthropic says it performs at the level of Claude Fable 5.1 on most work at lower cost.

Benchmarks · 11 reported

Benchmark	Score	Bar	Rank	Source
Terminal-Bench 4.0xhigh effort; SE ±2.6 pts; production safeguards enabled	66.4%		#2 of 10	Anthropic Opus 5.5 launch postSelf-reported
FrontierCode 1.1adaptive thinking, max effort	54.4%		#1 of 6	Anthropic Opus 5.5 launch postSelf-reported
CursorBench 4.0adaptive thinking, max effort	57.8%		–	Anthropic Opus 5.5 launch postSelf-reported
GDPval-AA v2.1knowledge work; max effort	1,846	Elo-style	–	Anthropic Opus 5.5 launch postSelf-reported
AutomationBenchpass rate; run and reported by Zapier during early access, no fallback models	40%		#2 of 8	Zapier, as cited in Anthropic Opus 5.5 launch post
Humanity’s Last Exam (with tools)with tools; max effort	67.7%		#1 of 16	Anthropic Opus 5.5 launch postSelf-reported
Terminal-Bench-Science 0.1agentic scientific research; SE ±3.5–5 pts	58.7%		–	Anthropic Opus 5.5 launch postSelf-reported
OSWorld 2.1partial-credit score; max effort	81.8%		–	Anthropic Opus 5.5 launch postSelf-reported
Chartographywith tools	89%		–	Anthropic Opus 5.5 launch postSelf-reported
Chartographyno tools; comparison column in a later post	64.4%		–	Anthropic Sonnet 5.5 launch postSelf-reported
AA-Briefcase v1.1knowledge work; comparison column in a later post	1,822	Elo-style	–	Anthropic Sonnet 5.5 launch postSelf-reported

Self-reported means the lab ran the test itself. A rank appears only where several labs report the same version of a benchmark.

Similar models

Models with the most benchmarks in common with Claude Opus 5.5, and the closest scores.

Compare 4 side by side

Terminal-Bench 4.0

#2 of 10

Long jobs in a real terminal

FrontierCode 1.1

#1 of 6

Patches good enough to merge

AutomationBench

#2 of 8

End-to-end business workflows

Humanity’s Last Exam (with tools)

#1 of 16

Expert questions, with search and code

Compare Claude Opus 5.5 side by side

Pre-filled with the two closest models. Swap any of them.

Open comparison

Sources

price, cache prices, batch, fast mode, tokenizer: platform.claude.com/docs/en/about-claude/pricing, checked 2 Oct 2026
context, max output, release, api id, knowledge cutoff, modalities, min cacheable prompt: platform.claude.com/docs/en/models/opus-5-5/overview, checked 2 Oct 2026
status, retirement commitment (not sooner than 2027-09-22): platform.claude.com/docs/en/about-claude/model-deprecations, checked 2 Oct 2026
refusal classifiers and refusal billing: platform.claude.com/docs/en/build-with-claude/refusals-and-fallback, checked 2 Oct 2026
benchmarks, positioning: anthropic.com/claude-opus-5-5, checked 2 Oct 2026
benchmarks (comparison column): anthropic.com/claude-sonnet-5-5, checked 2 Oct 2026
output speed: artificialanalysis.ai/models/claude-opus-5-5, checked 2 Oct 2026

Building on Claude Opus 5.5?

Get the architecture right before the bill arrives.

We size caching, routing and fallbacks for your workload on a free call, and tell you where a cheaper model is good enough.

Book a free call All AI models