Games / 3D / Coding
QQ Speed-inspired 3D racing demo
Inspect a creator-reported Opus 5.5 racing build through its public code, demo and original Chinese prompt.
riba2534 · Source reviewed · not reproduced
Anthropic / Model guide
Claude Opus 5.5
Explore source-linked coding projects, 3D worlds, research workflows and autonomous agents.
Model facts
- Context window
- 1M
- Max output
- 128K
- Input / 1M tokens
- $4
- Output / 1M tokens
- $20
Adaptive, always on · medium by default. Text and images in; text out. Reviewed 2026-09-25.
Official model documentationExamples to explore
All 14 examples
Games / 3D / Coding
Transport Ship: a browser FPS map
A creator-published shooter demo with its prompt, source and generation notes in the Opus 5.5 demo repository.
riba2534 · Source reviewed · not reproduced
Games / 3D / Coding
Pelican Bike: a browser 3D cycling scene
A source-linked Three.js cycling project with a public prompt, code and demo from riba2534.
riba2534 · Source reviewed · not reproduced
AI-generated illustration
Web UI / Coding
Improving load times across a web app
An Anthropic-reported web-app task evaluates performance improvements alongside behavior preservation.
Anthropic · Source reviewed · not reproduced
AI-generated illustration
Agents / Coding
Clio’s six-repository agent workflow
Clio reports a long-running Opus 5.5 task spanning six repositories in Anthropic’s launch material.
Clio, via Anthropic · Source reviewed · not reproduced
AI-generated illustration
Agents / Research
Column’s self-verifying subagents
Column reports improved delegation and external-document checks during an Opus 5.5 coding workflow.
Column, via Anthropic · Source reviewed · not reproduced
API pricing
$4 input · $20 output · $0.2 cache read per million tokens.
Standard API rates. Batch input/output is 50% lower; fast mode is $8 / $40. Cache writes: $5 (5 min), $8 (1 hour) per million tokens.
Official pricingReported benchmarks
| Benchmark | Claude Opus 5.5 |
|---|---|
| Terminal-Bench 4.0Opus 5.5: xhigh effort | 66.4% |
| FrontierCode v1.1Opus 5.5: max effort | 54.4% |
| CursorBench 4.0Opus 5.5: max effort | 57.8% |
| GDPval-AA v2.1Opus 5.5: max effort | 1846 Elo |
| AutomationBenchOpus 5.5: max effort | 40% |
| Humanity's Last ExamWith tools · Opus 5.5: max effort | 67.7% |
Vendor-reported launch results, September 22, 2026. Opus 5.5 uses max effort except Terminal-Bench (xhigh). Other models' effort, harness and tool settings can differ; see the original methodology. These are not AI IDE List tests and are separate from Anthropic's medium-effort cost charts.
Benchmark source and methodologyCompare Claude Opus 5.5
Claude Opus 5.5 vs Claude Opus 5
Opus 5.5 lowers standard input and output token prices by 20%. Use Opus 5 as a baseline and check for regressions on your existing tasks.
View comparisonClaude Opus 5.5 vs Claude Fable 5.1
Both offer a 1M context window and 128K output. Opus 5.5 has 60% lower standard input and output rates; the best choice still depends on task quality.
View comparisonClaude Opus 5.5 vs GPT-6 Astra
Opus 5.5 has lower standard token rates; Astra offers a slightly larger context window. Compare API tooling, long-context pricing and your own task results.
View comparison