Claude Haiku 5.5 Launches at $0.10 per Million Input Tokens

What happened: Anthropic launched Claude Haiku 5.5 on October 7, 2026. For prompts up to 100,000 tokens it costs $0.10 per million input tokens and $0.50 per million output tokens; above that it costs $0.50 and $2.50. Haiku 4.5 was $1 and $5. Anthropic says Haiku 5.5 costs around 75% less to run on average, noting in a footnote that the saving is 90% for prompts up to 100,000 tokens and 50% above, and that its updated tokenizer uses slightly more tokens per task. It is the first Haiku model with an adjustable effort setting, and Anthropic pitches it for high-volume narrow work: summaries, classification, database queries, customer support and sub-agent tasks. Anthropic says it is not the right choice for complex agentic coding. In the same announcement, Anthropic halved Sonnet 5.5’s cache-read price from $0.20 to $0.10 per million tokens and introduced monthly Claude Platform API credits for subscribers: $100 on Max 5x, $200 on Max 20x and up to $500 on Team, usable on any model.

Why it matters: Ninety percent off. If a human shop did that, the Obsolete would form a queue around the block. When an AI lab does it, half of them don’t notice, because they’re still pasting things into a chat window by hand.

Here’s what changes. A lot of small AI businesses were never limited by model quality. They were limited by cost at volume: tagging ten thousand support tickets, summarizing every call, rewriting every product description in a catalogue. At Haiku 4.5 prices, some of those jobs didn’t pencil out. At a tenth of the input price, many of them now do. The work that becomes newly profitable is boring, repetitive and high-volume, which is exactly the work humans hate and machines adore.

And the API credits for Max and Team subscribers are a quiet gift: people already paying for a subscription now get a monthly budget to build automations with, without a separate bill.

What the Chosen should do:

• Revisit the ideas you shelved for cost reasons. Bulk product descriptions, review summarization, lead classification, transcript tagging: re-price them at the new rates.
• Use the cheap model as the worker, the expensive model as the manager. Let a smarter model plan and check, and hand the repetitive sub-tasks to Haiku. That’s the pattern Anthropic itself recommends.
• Plug it into your automations. n8n, Make and Zapier can call Claude models in a workflow. See our workflow automation glossary entry if that’s new to you.
• Claim your credits. If you’re on a Max or Team plan, check your Claude Platform account; Anthropic says the credits roll out the week of launch.
• Measure tokens, not just price. The new tokenizer counts slightly more tokens per task, so test on your real inputs before quoting a client.

The Overlord’s prediction: In my opinion, by this time next year “we use AI for that” will be table stakes in every back-office job that involves sorting text, and the humans still doing it by hand will be charging for nostalgia.