Claude Haiku 5.5 launched on October 7, 2026, as Anthropic's third model in the Claude 5.5 family, following Opus 5.5 and Sonnet 5.5. It is live on the Claude Platform and in Claude Code, and also on AWS, Google Cloud and Microsoft Azure. The Claude Code v2.1.293 release notes published the same day confirm that Haiku 5.5 (model ID claude-haiku-5-5) is now the default Haiku model on the Anthropic API, with a 1M-token context window. The headline here is not raw capability but price: Anthropic says it costs around 75% less to run on average than Haiku 4.5.
Where the price actually lands
Haiku 5.5 pricing comes in two tiers, switched by prompt length:
| Prompt length | Input (per million tokens) | Output (per million tokens) |
|---|---|---|
| Up to 100K tokens | $0.10 | $0.50 |
| Over 100K tokens | $0.50 | $2.50 |
For comparison, Haiku 4.5 listed at $1 input and $5 output, so the short-prompt input rate is a tenth of the old one. Anthropic positions the model for high-volume small jobs such as classification, summarization and extraction, plus live support, voice agents and in-app assistants. It is also the first Haiku with built-in safeguards for a narrow set of high-risk cybersecurity requests, which the company says will not affect most everyday tasks.
The suggested role: doing the legwork for bigger models
Anthropic recommends using Haiku 5.5 as a subagent under Opus 5.5 or Sonnet 5.5, handling summarization, context compaction and database queries — the high-concurrency, cost-sensitive calls. The logic is straightforward: the main model reasons and decides, while swarms of cheap repetitive calls go to a model that costs a fraction as much. If you run long tasks through Claude Code cloud sessions, these sub-calls usually make up a visible share of the bill, so the switch matters more there than in single chats.
Subscription plans picked up API credit the same day
Two more pricing moves arrived alongside the launch. Cache reads on Claude Sonnet 5.5 were halved to $0.10 per million tokens, which helps any workload that rereads the same material. And Claude Max and Team plans now include monthly Claude Platform API credits: $100 for Max 5x, $200 for Max 20x, and up to $500 shared on Team, usable on any model, in your own code or a third-party harness. Part of the wall between subscription and API billing has come down.
Two caveats before you switch everything over
First, the tokenizer changed. Based on figures Anthropic shared with the press, the new tokenizer produces somewhat more tokens for the same text, and the 75% figure is an average-workload estimate — measure your own traffic before projecting savings. Second, 100K tokens is a hard line: once a prompt crosses it, rates jump fivefold. Before pointing batch document jobs at Haiku 5.5, check your prompt-length distribution so the bulk of calls do not land in the expensive tier.
