Claude Opus 5.5
Anthropic's current Opus and its recommended starting model, priced at $4/$20 per 1M tokens
Context window
1M
Input / 1M tokens
$4.00
Output / 1M tokens
$20.00
Provider
Anthropic
Cache reads $0.20 per 1M (0.05x input) and 5-minute cache writes $5 per 1M; Batch API is 50% off. Fast mode (up to 2.5x output speed) costs $8/$40 per 1M on the Claude API, Claude Models on Google Cloud (Anthropic managed) and Microsoft Foundry on Azure. The full 1M context is billed at standard rates. · Data verified 2026-11-11
Claude Opus 5.5 (September 22, 2026) is Anthropic's current Opus model and the one Anthropic recommends starting with 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 while costing 40% less to run than Opus 5 on typical workloads and generating output more than 30% faster. It has a 1M-token context window, 128K max output and adaptive thinking that is always on, with medium as the default effort. It replaced Claude Opus 5 at a lower price ($4/$20 vs $5/$25).
Capability index
Relative estimates (0-100) to place this model against its peers, grounded in published benchmarks.
How to access it
Available via the Claude API (claude-opus-5-5), Amazon Bedrock (anthropic.claude-opus-5-5), Claude Platform on AWS, Google Cloud, and Microsoft Foundry, and in the Claude apps and Claude Code on paid plans.
Strengths
- ✓Fable 5.1-level results on most work at less than half Fable 5.1's price, per Anthropic
- ✓1M token context window for long-running tasks
- ✓30%+ faster output generation than Opus 5
- ✓Anthropic says it is much less likely than recent models to take hard-to-reverse actions
- ✓Multimodal input (text and images)
- ✓Improved clarity in communication
Best for developers who...
When to choose it (and when not to)
Reach for Claude Opus 5.5 when...
- →When you need Fable-level performance at mid-tier pricing
- →For long-context coding and knowledge work
- →When cost-efficiency matters alongside capability
Look elsewhere if...
- ✕You need the strongest Claude results on long-horizon agentic work (use Fable 5.2, at twice the price)
- ✕For simple, quick queries where faster/cheaper models suffice
How to use it
- ›Use the 1M context window for comprehensive documents and conversation history
- ›Adaptive thinking is always on with medium as the default effort; raise effort to high or xhigh for hard coding and agentic work
- ›Thinking cannot be disabled and forced tool_choice (any or tool) returns a 400; use tool_choice auto with clear instructions instead
Quickstart
Pythonimport anthropic
client = anthropic.Anthropic()
message = client.messages.create(
model="claude-opus-5-5",
max_tokens=1024,
messages=[{"role": "user", "content": "Your prompt here"}]
)
print(message.content[0].text)Model ID is 'claude-opus-5-5' across all platforms (API, Bedrock, GCP, Azure)
API model id: claude-opus-5-5
Benchmarks
| Benchmark | Score | Notes |
|---|---|---|
| Terminal-Bench 4.0 | 66.4% (xhigh effort) | Anthropic-reported; 64.8% at max effort |
| Anthropic Automated Behavioral Audit | Strongest-performing model tested | Comprehensive alignment and safety evaluation |
| SWE-bench Pro | 89.9% (max effort) | Anthropic system card, average of 5 trials |
| DeepSWE v1.1 | 74.2% (max effort) | Anthropic system card, average of 5 trials |
| OSWorld 2.1 | 81.8% (max effort) | Anthropic-reported, full 108-task set, partial score |
Source: Anthropic System Card
Compare Claude Opus 5.5
Claude Opus 5.5 vs GPT-6.1 Sol
OpenAI - 1,050,000 ctx
Claude Opus 5.5 vs Gemini 4 Argon
Google - 1M (shared input + output) ctx
Compare Claude Opus 5.5 with any other model
Build a comparison →All model comparisons →