Claude Opus 5.5
Anthropic's flagship model for long-running agentic coding and knowledge work — $4 / $20 per million tokens, 1M context.
Claude Opus 5.5 launched on September 22, 2026. It scores 58 on the Artificial Analysis Intelligence Index at max effort, supports a 1M-token context window with 128K output, and always reasons with adaptive thinking. This page collects its pricing, specs, benchmarks and every head-to-head comparison, plus a calculator for estimating your monthly token cost.
Last updated 2026-10-01 · Data from Anthropic and Artificial Analysis
Claude Opus 5.5 Key Specifications
| Spec | Claude Opus 5.5 |
|---|---|
| Developer | Anthropic |
| API model ID | claude-opus-5-5 |
| Released | Sep 22, 2026 |
| Context window | 1M tokens |
| Max output | 128K tokens (300K on Batch API) |
| Input / output price | $4 / $20 per MTok |
| Knowledge cutoff | Jun 2026 |
| Thinking | Adaptive, always on, cannot be disabled |
| Default effort | medium |
| Input / output | Text + images in, Text out |
| Platforms | Claude API, Bedrock, Google Cloud, Microsoft Foundry, Claude Platform on AWS |
| Retirement | Not sooner than Sep 22, 2027 |
Claude Opus 5.5 Pricing
Claude Opus 5.5 costs $4 per million input tokens and $20 per million output tokens. Anthropic says that is 20% less than Opus 5 for input and output and 60% less for cache reads.
| Tier | Price per MTok |
|---|---|
| Standard API | $4 input / $20 output |
| Batch API (50% off) | $2 input / $10 output |
| Cache write (5 min) | $5 |
| Cache write (1 hour) | $8 |
| Cache read | $0.20 |
| Fast mode (research preview) | $8 / $40 per MTok, up to 2.5x speed |
Per-token price is not the same as cost per task. Artificial Analysis measures about $5.98 per Intelligence Index task at max effort, and far less at the default medium effort. See Opus 5.5 vs Sonnet 5 and Opus 5.5 vs Sonnet 5.5 for per-task comparisons.
Claude Opus 5.5 Cost Calculator
Estimate your monthly Claude Opus 5.5 bill from your own request volume and token counts.
Estimated monthly cost (30 days)
$540.00
Average cost per request
$0.02
Token-price arithmetic only, using Claude Opus 5.5 list prices. Cache writes, fast mode and tool or search surcharges are not included, and the Batch estimate ignores prompt caching. Real cost per task also depends on how many reasoning tokens the model uses, so measure on your own workload.
Claude Opus 5.5 Benchmarks & Speed
Independent measurements from Artificial Analysis (Intelligence Index v4.3.2):
| Metric | Result | Setting |
|---|---|---|
| Intelligence Index v4.3.2 | 58 | Max effort |
| Intelligence Index v4.3.2 | 51 | Medium effort (default) |
| Cost per Intelligence Index task | $5.98 | Max effort |
| Cost to run the full Intelligence Index | ~$1,627 | Medium effort |
| Output tokens per task | ~26k | Medium effort |
| Output speed | ~72 tokens/s | Medium effort |
Benchmarks published by Anthropic in the Opus 5.5 announcement:
| Benchmark | Score | Note |
|---|---|---|
| Terminal-Bench 4.0 | 66.4% | Agentic coding; xhigh effort |
| FrontierCode 1.1 Main | 54.4% | Coding |
| CursorBench 4.0 | 57.8% | Coding |
| OSWorld 2.1 | 81.8% | Computer use |
| GDPval-AA v2.1 | 1846 Elo | Knowledge work |
| AA-Briefcase v1.1 | 1822 Elo | Business reasoning |
| Humanity's Last Exam | 67.7% | With tools |
| Chartography | 64.4% | Visual recognition; no tools |
What Claude Opus 5.5 Is Best For
Agentic coding
Long-running coding agents and codebase-wide migrations or audits.
Computer use
Operating desktop and browser software from screenshots.
Knowledge work
Financial analysis, business workflows and research tasks.
Hard reasoning
Problems where answer quality matters more than per-token price.
Claude Opus 5.5 vs Other Models
Claude Opus 5.5 vs Claude Sonnet 5.5
Opus leads 7 of 8 benchmarks; Sonnet wins Terminal-Bench, is faster and costs half as much.
Claude Opus 5.5 vs Claude Fable 5.1
Opus is 60% cheaper, faster and scores higher on all seven benchmarks.
Claude Opus 5.5 vs GPT-6 Astra
Opus leads 7 of 9 and costs 60% less; Astra edges ahead on math and scientific reasoning.
Claude Opus 5.5 vs GPT-6 Sol
Opus medium outscores Sol max; Sol is 50% cheaper per token but hits a lower quality ceiling.
Claude Opus 5.5 vs Claude Sonnet 5
Opus leads all 9 benchmarks and costs 4.3× less to evaluate at default effort despite 2× higher per-token pricing.
Claude Opus 5.5: FAQ
How much does Claude Opus 5.5 cost?
Claude Opus 5.5 costs $4 per million input tokens and $20 per million output tokens on the Claude API. Batch requests are half price ($2 / $10), cache reads cost $0.20 per million tokens, and cache writes cost $5 (5 minutes) or $8 (1 hour). Fast mode, a research preview, costs $8 / $40 per million tokens for up to 2.5x faster output. A typical request with 100K input and 20K output tokens costs about $0.80.
Can I use Claude Opus 5.5 for free?
No. As of September 2026, Opus 5.5 is not available on the free plan at claude.ai. It requires a Pro, Max, Team or Enterprise subscription, or pay-as-you-go access through the Claude API, Amazon Bedrock, Google Cloud or Microsoft Foundry.
Is Claude Opus 5.5 available in Claude Code?
Anthropic's announcement references Claude Code alongside Opus 5.5's focus on long-running agentic coding, and customer quotes describe using it there. It does not name a default model for Claude Code, so check your Claude Code model setting.
Does Claude Opus 5.5 support video?
No video input or output is listed in Anthropic's model documentation. Opus 5.5 accepts text and images as input and returns text only.
What is the context window of Claude Opus 5.5?
Claude Opus 5.5 has a 1M-token context window at standard per-token pricing, and a maximum output of 128K tokens (300K on the Batch API).
How fast is Claude Opus 5.5?
Artificial Analysis measures about 72 output tokens per second at medium effort. Anthropic lists latency as moderate, and fast mode offers up to 2.5x faster output at $8 / $40 per million tokens.
How does Claude Opus 5.5 score on benchmarks?
On the Artificial Analysis Intelligence Index v4.3.2 it scores 58 at max effort and 51 at medium effort. In Anthropic's published table it reaches 66.4% on Terminal-Bench 4.0, 81.8% on OSWorld 2.1 and 67.7% on Humanity's Last Exam with tools.
Can Claude Opus 5.5 turn thinking off?
No. Opus 5.5 uses adaptive thinking that is always on and cannot be disabled. You control how much it reasons with the effort setting, which defaults to medium.
What is Claude Opus 5.5's knowledge cutoff and when will it retire?
The knowledge cutoff is June 2026. Anthropic guarantees availability until at least September 22, 2027.