Skip to content
All systems operational0 AI providers monitored, polled every 2 minutes
Live status

Claude Opus 5.5

Flagship

by Anthropic

Claude Opus 5.5 shipped on September 22, 2026 as claude-opus-5-5, and it is the rare release that is both cheaper and better than the model it follows. Pricing is $4 per million input tokens and $20 output, 20 percent under Claude Opus 5 at $5 and $25, and cache reads fall to $0.20, priced at 0.05x base input where every other Claude model uses 0.1x, which is a 60 percent cut on the line that dominates long agent sessions. Cache writes are $5 for five minutes and $8 for an hour, Batch halves everything to $2 and $10, and fast mode on the Claude API runs at $8 and $40. The context window is 1 million tokens with no long-context surcharge and 128K output. Anthropic's own framing is that Opus 5.5 performs at the level of Claude Fable 5.1 on most work while costing about 40 percent less to run than Opus 5, and the independent boards agree more than launch claims usually earn: it is first on the Vals Index at 69.69 percent, ahead of Fable 5.1 at 68.83 and GPT-6 Astra at 66.61, first on the Artificial Analysis Intelligence Index at 58 against 53 for Astra and Fable 5.1, and first on Cognition's FrontierCode board at 54.6 percent. Anthropic's table adds 66.4 percent on Terminal-Bench 4.0 at xhigh effort, where Astra posts 57.9, and 67.7 percent on Humanity's Last Exam with tools; Vals measured Terminal-Bench 4.0 independently at 61.62. Anthropic no longer reports SWE-bench Verified, MMLU-Pro, or GPQA Diamond, and Vals archived all three as saturated on September 1, so those columns stay empty here. Four API changes will break code written for Opus 5: thinking is always on and cannot be disabled, the default effort drops to medium from high, forced tool choice returns a 400, and computer use requires the computer_toolset_20260801 tool. It is also the first Opus to ship with Fable 5.1-class safeguards, which reroute most cybersecurity tasks to Opus 4.8. Claude Code now defaults to Opus 5.5 on Pro, Max, Team, Enterprise, and API accounts, and it is available on the Claude API, Amazon Bedrock, Google Cloud, and Microsoft Foundry. Opus 5 is not deprecated and remains available at its old rate.

Input Price

$4.00

per 1M tokens

Output Price

$20.00

per 1M tokens

Context Window

1M

tokens

Released

2026-09

API access

Capabilities

textvisiontool-usecodereasoning

Key Strengths

  • ✓$4/$20 per 1M tokens, 20 percent under Opus 5
  • ✓Cache reads at $0.20, a 0.05x multiplier against the usual 0.1x
  • ✓First on the Vals Index at 69.69 percent and on Artificial Analysis at 58
  • ✓FrontierCode v1.1 at 54.6 percent, top of Cognition's board
  • ✓1M token context with no long-context surcharge, 128K output
  • ✓Terminal-Bench 4.0 at 61.62 percent measured by Vals

Best For

  • ▸Long-running agentic coding and codebase-wide migrations
  • ▸Computer use and GUI automation
  • ▸Knowledge work and financial analysis at below Fable 5.1 cost
  • ▸Agent sessions that replay a large cached prefix every turn

Benchmark Scores

BenchmarkScoreDescription
FrontierCode v1.154.6Agentic coding on frontier software engineering tasks (Main split)
Terminal-Bench 4.061.6Long-horizon agentic work in a terminal across software, science, ML, operations, hardware, security, and media (66 tasks, all-or-nothing verifiers)
Humanity's Last Exam (tools)67.7Multidisciplinary expert-level reasoning with tool access

Scores sourced from public benchmark datasets. See full benchmark leaderboard for all models.

Pricing Details

Input tokens

$4.00

per 1M tokens

Output tokens

$20.00

per 1M tokens

Estimated cost per 1K requests

$14.00

~1K input + ~500 output tokens avg

Prices are subject to change. Check the official documentation for current pricing. See the cost calculator for detailed estimates.

Related Models

View DocumentationCompare ModelsCost CalculatorFull Pricing Guide