Claude Haiku 5.5 arrives on Vercel's AI Gateway with new effort levels
Vercel has made Anthropic's Claude Haiku 5.5 available on its AI Gateway. According to Vercel, it is the first Haiku model with effort levels and the fastest Claude model at standard speed.
- Claude Haiku 5.5 from Anthropic is now available on Vercel's AI Gateway.
- Vercel calls it the fastest Claude model at standard speed, built for high-volume, cost-sensitive tasks.
- It is the first Haiku model with effort levels: low, medium, high, xhigh and max.
- Thinking can be off at low, medium and high, but must stay on at xhigh and max.
- AI Gateway reflects provider pricing with no markup and supports Zero Data Retention.
What Haiku 5.5 is built for
Vercel describes Claude Haiku 5.5 as a model for high-volume, cost-sensitive tasks. The examples it lists are summaries, context compaction, database queries and classification. It can also act as a subagent alongside larger Claude models on coding work.
Vercel says it is the fastest Claude model at standard speed, which it argues suits live customer support and browser use. Teams already running Claude Haiku 4.5 are advised to test Haiku 5.5 as a direct upgrade.
Effort levels and adaptive thinking
Haiku 5.5 is the first Haiku model with effort levels. It uses adaptive thinking and supports five settings: low, medium, high, xhigh and max. These control how much the model thinks and how many tokens it uses.
- Thinking can be turned off at low, medium and high.
- At xhigh and max, thinking must stay on.
Vercel also notes that the model supports Zero Data Retention on AI Gateway. It ships with biology and cybersecurity safeguards, so it may decline some requests in those areas.
How to call it
Developers can reach the model through the AI SDK, the Chat Completions, Responses and Anthropic Messages APIs, or a coding agent connected to AI Gateway. The model name is anthropic/claude-haiku-5.5. Effort is set with reasoning in the AI SDK or reasoning_effort in Chat Completions.
Vercel's example uses the AI SDK's streamText function with the reasoning level set to low, asking the model to extract a city and date from a short sentence.
To use the model in Claude Code, Codex, Cursor and other agents, Vercel says to install the latest Vercel CLI with npm i -g vercel@latest, run vercel ai-gateway setup, and then select anthropic/claude-haiku-5.5 in the agent. A model playground is also available for trying it out.
What AI Gateway adds
AI Gateway offers a unified API for calling models, tracking usage and cost, and configuring retries, failover and performance optimizations. Vercel says this is aimed at higher-than-provider uptime. It also includes custom reporting, Zero Data Retention support, budgets for API keys and routing rules.
On pricing, Vercel states that AI Gateway reflects provider pricing with no markup. It does not charge a platform fee on inference, including on Bring Your Own Key (BYOK) requests. The source does not give per-token prices for Haiku 5.5.
For teams that already use Haiku 4.5 for classification, summarization or support workloads, the release offers a model with adjustable reasoning effort through the same gateway they may already use.
Vercel: Claude Haiku 5.5 now available on AI Gateway
