Claude Haiku 5.5's price cut, and what to do about it

Anthropic launched a new version of its cheapest, fastest model, Claude Haiku 5.5, on October 7, 2026, with API pricing cut by up to 90% compared to the model it replaces. For prompts under 100,000 tokens, input tokens dropped from $1.00 to $0.10 per million and output tokens dropped from $5.00 to $0.50 per million. For prompts over 100,000 tokens, the price is $0.50 input and $2.50 output per million, still a 50% cut from before. Anthropic describes the blended average saving across typical use as around 75%.

Haiku isn't Anthropic's most capable model (that's Opus), it's the fast, cheap tier built for high-volume work that doesn't need heavy reasoning: sorting support tickets, pulling fields off a form, tagging a batch of records. That's exactly the kind of task a small business is most likely to run thousands of times a day if the price is low enough, so a 90% cut there moves more of those tasks from 'too expensive to bother automating' to 'cheap enough to just run.'

Alongside the Haiku 5.5 launch, Anthropic started rolling out free monthly API credit to paid Claude subscribers the week of October 7, 2026: $100 a month for Claude Max 5x subscribers, $200 a month for Max 20x subscribers, and up to $500 a month for Team plans, pooled across everyone on the team. The credit can be spent on calls to any Claude model through Anthropic's API, not only Haiku, though it does not cover interactive use of Claude Code itself.

If your business already pays for a Claude Max or Team seat, this isn't something you sign up for separately. It should land on the account automatically as the rollout reaches you. Check your plan's billing or API console this week to see whether it has shown up, and what it's actually worth to you, before deciding whether to build anything new against it.

Not directly. The $100 to $500 figures above, and the Haiku price cut, apply to Anthropic's API, the metered, pay-as-you-go usage that powers custom tools, chatbots, and automations, not the flat monthly seat price for chatting inside Claude.ai or Claude Code. If nobody at your company has ever called Claude's API directly, your monthly bill doesn't change because of this announcement.

It still matters indirectly if you pay a vendor or an in-house developer to build something on Claude: a website chatbot, an inbox triage tool, a document processor. Ask whoever maintains it which model it calls. If it's already on a Haiku model, it just got cheaper to run with no code change needed. If it was built on a pricier model for a task that doesn't need deep reasoning, this is the moment to ask whether swapping to Haiku 5.5 would hold up on quality.

Three concrete moves. First, if you're on Claude Max or Team, confirm the new API credit has reached your account and decide whether anything is worth building with that free usage attached. Second, if you already have an automation or feature running on an older Haiku model, ask whoever built it to test Haiku 5.5 on a sample of real requests before switching the production default, since the swap is usually a one-line model name change, not a rebuild. Third, if a repetitive task has been sitting on a 'too expensive to automate' list, this is the price point worth checking it against again.