Claude Opus 5.5 Release: Specs and Pricing
Anthropic released Claude Opus 5.5 on September 22, 2026. It features a 1M token context window, 128k max output, and $4/$20 pricing for agentic coding tasks.
Official Announcement and Specifications
Anthropic officially launched Claude Opus 5.5 (claude-opus-5-5) on September 22, 2026. The announcement describes the model as designed for long-running agentic coding and knowledge work. According to the official release notes, the model operates with a default context window of 1,000,000 tokens. The maximum output limit is set at 128,000 tokens. The system includes an always-on adaptive thinking feature.
- Default context window: 1,000,000 tokens
- Maximum output tokens: 128,000
- Input modalities: Text, Image
- Output modality: Text
Pricing Structure and Availability
The release notes specify the pricing for Claude Opus 5.5 at $4 USD per million input tokens and $20 USD per million output tokens. For comparison, the previous iteration, Claude Opus 5, was listed at $5 USD per million input tokens and $25 USD per million output tokens. The source explicitly states these figures but does not attribute the price change to specific architectural changes or performance improvements.
- Claude Opus 5.5 Input Price: $4 USD per MTok
- Claude Opus 5.5 Output Price: $20 USD per MTok
- Claude Opus 5 Input Price: $5 USD per MTok
- Claude Opus 5 Output Price: $25 USD per MTok
Evaluation Procedure for Engineers
Engineers should evaluate the model based on its documented specifications rather than inferred capabilities. A practical evaluation procedure involves testing the model against specific agentic coding tasks to measure performance. First, test the model using a representative dataset that requires long-context processing. Measure the latency and accuracy of the responses. Compare the results against baseline metrics from previous models if available in your internal benchmarks. Validate that the 1M token context window functions as described by inputting large documents and checking for truncation errors.
Artificial Analysis benchmark review
Benchmark review updated on . These scores apply to the linked artifact and recorded methodology; they do not establish the performance of a different runtime or quantized model.
| Measurement | Recorded result | Measured |
|---|---|---|
| Artificial Analysis Intelligence Index (v4.3) ↗ | 57.6 index_points | 2026-09-22 |
Review the full benchmark provenance and source methodology →
Primary sources
Frequently asked questions
What is the exact date of the Claude Opus 5.5 release?
The official announcement date for Claude Opus 5.5 is September 22, 2026. This date is recorded in the event metadata and corresponds to the publication of the release notes.
Does the source confirm that local installation is possible?
No. The source states that open weights are not confirmed. It does not establish whether local installation is impossible or possible, nor does it mandate API access as the only method. The availability statements refer to cloud platforms and APIs.