Claude Haiku 5.5 now available on AI Gateway
By Steven Van ·
Anthropic’s model adds five thinking effort levels and supports Zero Data Retention for high-volume, cost-sensitive tasks.
High-volume tasks such as summaries, context compaction, database queries and classification are the focus of Anthropic’s Claude Haiku 5.5, now available through Vercel AI Gateway. Vercel describes it as the fastest Claude model at standard speed, suited to live customer support and browser use, and suggests testing it as a direct upgrade from Haiku 4.5. It can also work alongside larger Claude models as a coding subagent.
Adjust thinking effort
Haiku 5.5 is the first Haiku model with effort levels. Its adaptive thinking supports low, medium, high, xhigh and max, controlling how much the model thinks and how many tokens it uses. Thinking can be switched off at low, medium and high, but must remain on at xhigh and max.
Connect apps and coding agents
Use the model name anthropic/claude-haiku-5.5 through the AI SDK, Chat Completions, Responses or Anthropic Messages APIs. Set effort with reasoning in the AI SDK or reasoning_effort in Chat Completions. The Responses API documentation explains how to point the OpenAI SDK at AI Gateway’s base URL and authenticate requests.
For Claude Code, Codex, Cursor and other coding agents, install the latest Vercel CLI, run vercel ai-gateway setup, then select anthropic/claude-haiku-5.5 in the agent.
Data retention and safeguards
Haiku 5.5 supports Zero Data Retention on AI Gateway. It also includes biology and cybersecurity safeguards, so it may decline some requests in those areas.