Claude Haiku 5.5 now available on AI Gateway
Anthropic’s Claude Haiku 5.5 is now accessible via Vercel’s AI Gateway, offering faster, cost-efficient subagent capabilities with adjustable reasoning effort levels and zero data retention.
Claude Haiku 5.5 from Anthropic is now available on Vercel’s AI Gateway, designed for high-volume, cost-sensitive tasks such as summaries, context compaction, database queries, and classification. It functions as a subagent alongside larger Claude models for coding work and is positioned as the fastest standard-speed model, making it suitable for live customer support and browser use. Users can test it as a direct upgrade from Claude Haiku 4.5 where the latter is currently deployed.
The new model introduces effort levels—low, medium, high, xhigh, and max—controlling how much the model thinks and the number of tokens it consumes. Thinking can be disabled at low, medium, and high levels, while xhigh and max require it to remain on. Haiku 5.5 also supports Zero Data Retention on AI Gateway and includes built-in safeguards for biology and cybersecurity, which may result in declined requests in those domains.
Claude Haiku 5.5 can be accessed through the AI SDK, Chat Completions, Responses, and Anthropic Messages APIs, or via a coding agent connected to AI Gateway. Effort levels are set using reasoning in the AI SDK or reasoning_effort in Chat Completions. For use in coding agents like Claude Code, Codex, and Cursor, users must install the latest Vercel CLI, run setup, and select anthropic/claude-haiku-5.5 in the agent configuration.
AI Gateway provides a unified API for model access, usage tracking, cost management, and performance optimizations such as retries, failover, and higher uptime. It includes features like Zero Data Retention support, custom reporting, API key budgets, and routing rules. Pricing reflects provider rates with no markup, and no platform fee is charged on inference, including Bring Your Own Key requests.