MiniMax M3.1 Flash
A multimodal model for coding assistants, computer interaction and workflows that require planning and tool calls.
Claude Haiku 5.5
www.anthropic.com
A hosted model for high-volume summarization, classification, extraction, customer support and focused coding workflows.
Visit website
Pricing
Freemium
Available on
Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry, Claude Platform on AWS, Web, iOS, Android, Claude Code
Category
Llm Models
What is Claude Haiku 5.5?
Claude Haiku 5.5 is Anthropic's small model for high-volume, cost-sensitive work. Its documented tasks include summaries, classification, database queries, customer support, browser use and focused coding subagent work. Anthropic positions larger models as better suited to complex agentic coding.
Teams can access Haiku 5.5 through Claude's apps, Claude Code, the direct API and supported cloud platforms. It can handle supporting tasks within larger workflows, including compacting conversation history.
Key features
Text and image understanding
Accepts text and image inputs and produces text, supporting workflows that combine written material with visual context.
Large context and output allowances
Provides a 1-million-token context window and a standard maximum output of 128,000 tokens.
Adaptive thinking
Thinking is enabled by default at medium effort. Developers can adjust effort to control reasoning depth.
Focused agent tasks
Supports browser-use workflows and coding subagent assignments, alongside summaries and compaction for larger systems.
Batch processing
Batch processing reduces API input and output token prices by 50%, offering a lower-cost route for eligible asynchronous work.
Who is it for?
Support conversation summaries
Support teams can turn lengthy conversations into concise handoff notes.
Request classification
Application developers can classify incoming requests before routing them to a workflow or specialist.
Focused coding assignments
Engineering teams can assign a narrow edit or investigation as one part of a larger development workflow.
How to use Claude Haiku 5.5
- Choose Claude's apps, the direct API or a supported cloud platform.
- For direct API access, create a Claude Console account and configure an API key or workload identity federation.
- Try a representative task in the Console playground, then select claude-haiku-5-5 in your integration.
- Set effort and output allowances, and check migration requirements before reusing an older Haiku request configuration.
Claude Haiku 5.5 pricing
Freemium
Haiku 5.5 is selectable on Claude's Free and paid plans. Direct API access uses token-based billing, with higher per-token prices when prompts exceed 100,000 tokens. Partner clouds have separate pricing.
Direct API — prompts up to 100,000 tokens
USD 0.10 per million input tokens; USD 0.50 per million output tokens
Usage-based per request; no monthly interval
Standard global direct API rates. Prompt length includes cache reads and writes when determining the pricing tier.
Direct API — prompts over 100,000 tokens
USD 0.50 per million input tokens; USD 2.50 per million output tokens
Usage-based per request; no monthly interval
Standard global direct API rates. The higher tier applies to the request when its prompt exceeds the threshold.
What to know before you use it
- Migration requires removing temperature, top_p and top_k and replacing assistant-prefill requests. Structured outputs are unavailable on Amazon Bedrock, and Priority Tier is unsupported.
- The newer tokenizer counts approximately 30% more tokens for the same text than Haiku 4.5. Thinking also consumes the output allowance, affecting prompt budgets and response sizing.
- Cybersecurity safeguards block penetration testing and certain techniques associated with attackers, limiting suitability for some security workflows.
Frequently asked questions
Can I select Haiku 5.5 on Claude's Free plan?
Yes. Anthropic lists Haiku 5.5 as selectable for Free, Pro, Max, Team and Enterprise users on web, iOS and Android.
What model ID does the direct API use?
Use claude-haiku-5-5. It is a fixed model ID without a date suffix or separate alias.
Can stored thinking blocks be reused across unrelated accounts?
Thinking blocks work only in the account that produced them or accounts linked to it.