Technical Specifications of Claude Haiku 5.5
| Specification | Claude Haiku 5.5 |
|---|---|
| Model ID | claude-haiku-5-5 |
| Provider | Anthropic |
| Model type | Text generation / multimodal understanding |
| Input | Text, images |
| Output | Text |
| Context window | 1M tokens |
| Maximum output | 128K tokens |
| Batch maximum output | 300K tokens with the documented beta configuration |
| Thinking | Adaptive |
| Effort levels | low, medium, high, xhigh, max |
| Default effort | medium |
| Knowledge cutoff | June 2026 |
| Released | October 7, 2026 |
| Latency profile | Fastest in Anthropic's current comparison |
What Is Claude Haiku 5.5?
Claude Haiku 5.5 is Anthropic's high-volume, latency-sensitive model for classification, extraction, routing, customer support, summaries, database queries, and subagent workloads. It is also capable of coding, tool use, browser/computer workflows, and longer reasoning tasks.
Compared with Claude Haiku 4.5, it expands the context window from 200K to 1M tokens, increases maximum output from 64K to 128K, and introduces adaptive thinking with configurable effort.
Main Features of Claude Haiku 5.5
- 1M-token context: Suitable for large documents, long conversations, codebases, and retrieval-heavy workflows.
- Adaptive thinking: The model dynamically decides how much reasoning to use; developers control the trade-off with
effort. - Five effort levels:
low,medium,high,xhigh, andmax, withmediumas the default. - Very low latency: Anthropic positions Haiku 5.5 as its fastest model, making it appropriate for high-volume and real-time workloads.
- Multimodal input: Supports text and image inputs with text output.
- Agentic/tool workflows: Suitable for subagents and supported browser/computer-use workflows.
- Large output capacity: Up to 128K output tokens on the standard API and up to 300K in the documented Batch API beta configuration.
Benchmark Performance of Claude Haiku 5.5
Anthropic's October 2026 evaluations report the following results:
| Benchmark | Haiku 5.5 | Haiku 4.5 | Sonnet 5.5 |
|---|---|---|---|
| GDPval-AA v2.1 | 1620 | 735 | 1840 |
| AA-Briefcase v1.1 | 1578 | 614 | 1824 |
| OSWorld 2.1, offline | 72.4% | 15.7% | 83.9% |
| Humanity's Last Exam, no tools | 45.9% | 10.2% | 56.9% |
| Humanity's Last Exam, with tools | 57.4% | 18.7% | 64.5% |
| Terminal-Bench 4.0 | 39.2% | 0.0% | 70.6% |
| FrontierCode 1.1 Main | 46.4% | — | 52.1% |
The results show a major improvement over Haiku 4.5, while Sonnet 5.5 remains the stronger choice for workloads where maximum reasoning and coding capability outweigh latency and cost.
Claude Haiku 5.5 vs Haiku 4.5 vs Sonnet 5.5
| Model | Context | Max output | Latency | Best fit |
|---|---|---|---|---|
| Claude Haiku 5.5 | 1M | 128K | Fastest | High-volume automation and latency-sensitive agents |
| Claude Haiku 4.5 | 200K | 64K | Very fast | Existing legacy integrations |
| Claude Sonnet 5.5 | 1M | 128K | Fast | Higher-end coding and knowledge work |
Haiku 5.5 is the better starting point for new high-volume applications. Sonnet 5.5 is preferable when the application requires stronger performance on difficult coding, reasoning, or knowledge-work tasks.
Limitations and Migration Considerations
Developers migrating from Haiku 4.5 should account for several breaking changes. Manual extended thinking with thinking: {"type":"enabled","budget_tokens":...} is no longer supported; adaptive thinking and output_config.effort should be used instead.
Non-default temperature, top_p, and top_k values return a 400 error. Because adaptive thinking is enabled by default, responses can begin with thinking blocks, so clients should select content by type rather than assuming the first block is the final answer.
Haiku 5.5 also uses the newer tokenizer used by Claude 4.7 and later. Anthropic says the same text produces approximately 30% more tokens than with Haiku 4.5, so existing token budgets and cost estimates should be recalculated.
Representative Use Cases
- Classification and routing: High-volume intent detection, ticket routing, and workflow dispatch.
- Information extraction: Extracting entities, fields, summaries, and structured information from large document collections.
- Customer support: Fast conversational responses and tool-assisted support workflows.
- Agentic automation: Low-latency subagent calls for repetitive research, verification, and execution tasks.
- Coding assistance: Code generation, review, debugging, and agentic development; Terminal-Bench results show a large improvement over Haiku 4.5.
- Long-context analysis: Large documents, long histories, and repository-scale context without repeatedly compressing input.
How to Access Claude Haiku 5.5 API through CometAPI
Step 1: Sign Up for a CometAPI Account
Create or sign in to your CometAPI account. CometAPI provides a unified API platform for models from Anthropic, OpenAI, Google, and other providers, allowing developers to manage model access through a single API interface and API key.
Claude Haiku 5.5 is currently listed in the CometAPI model catalog as an Anthropic text-generation model, although its CometAPI availability is currently marked Coming soon.
Step 2: Select Claude Haiku 5.5
The official Anthropic model ID is:
claude-haiku-5-5
Claude Haiku 5.5 is designed for high-volume, cost-sensitive workloads where speed and efficiency are important. Anthropic describes it as its fastest model to date and positions it for tasks such as summarization, classification, routing, real-time support, browser use, and subagent workloads.
Once CometAPI enables the model for API access, developers can use claude-haiku-5-5 through the CometAPI model interface.
Step 3: Send a Request through the CometAPI API
CometAPI provides a unified API layer for accessing models from different providers. Where to call it: Anthropic Messages format and Chat format.
Insert your question or request into the content field—this is what the model will respond to . Process the API response to get the generated answer.