Anthropic launches Claude Haiku 5.5, cuts API prices by up to 90%
Representative image: yatsusimnetcojp / Pixabay
Anthropic released Claude Haiku 5.5 on October 7, 2026, pricing short prompts at $0.10 per million input tokens, up to 90% below Haiku 4.5.
What Happened
Anthropic released Claude Haiku 5.5, the newest version of its smallest and cheapest AI model, on October 7, 2026, and cut the price developers pay to use it through its API by up to 90 percent. For prompts of up to 100,000 tokens, the model costs $0.10 per million input tokens and $0.50 per million output tokens, against $1 and $5 for the previous version, Haiku 4.5. The model accepts a context of up to one million tokens. With the change, Anthropic's entry price matches that of OpenAI's small GPT-6 Luna model.
Key Facts
- Release date: October 7, 2026.
- Short-prompt price (up to 100,000 tokens): $0.10 per million input tokens and $0.50 per million output tokens.
- Previous model, Haiku 4.5: $1 input and $5 output per million tokens, so the short-prompt rate is 90 percent lower.
- Long-prompt price (above 100,000 tokens): $0.50 input and $2.50 output per million tokens, still half the old rate.
- Average saving: Anthropic estimates running costs fall by about 75 percent on average once its updated tokenizer is taken into account.
- Batch requests are reported to get a further 50 percent discount.
- In the same announcement, the cache-read price for the larger Sonnet 5.5 model was halved to $0.10 per million tokens.
Why It Matters
Small models do most of the everyday work in AI products: sorting support tickets, summarising documents, pulling data out of forms, routing requests and powering chat features inside apps. For a company running millions of such calls a day, the per-token price decides whether a feature is affordable at all. A tenfold cut on the most common size of request changes that sum sharply.
The move also shows how fast prices are falling at the low end of the market. Anthropic set its entry price level with OpenAI's GPT-6 Luna, although the two differ on what they charge for longer prompts. Reports describe the range of savings as anywhere from 75 to 90 percent, and the difference matters. The 90 percent figure applies to the short-prompt list price. The 75 percent figure is the company's estimate of the average saving across real workloads, because the new tokenizer splits text differently and can change how many tokens a task uses.
For India, where a large share of AI products are built by start-ups and IT services firms serving price-sensitive customers, cheaper small models lower the cost of building in local-language support, customer service and back-office automation. Developers paying in rupees for dollar-priced APIs feel price cuts of this size directly in their margins.
| Item | Haiku 4.5 | Haiku 5.5 |
|---|---|---|
| Input, short prompts (per million tokens) | $1.00 | $0.10 |
| Output, short prompts (per million tokens) | $5.00 | $0.50 |
| Input, prompts above 100,000 tokens | not stated | $0.50 |
| Output, prompts above 100,000 tokens | not stated | $2.50 |
Impact
Short-term: Developers already using Haiku can cut their bills sharply by switching, and teams comparing small models now face matching entry prices from the two biggest US AI labs.
Long-term: Falling prices for capable small models make it cheaper to put AI into high-volume, low-margin tasks, which could widen use well beyond chat assistants. It also puts pressure on rivals to cut their own small-model prices.
Who is affected: Software developers, AI start-ups, IT services companies, and businesses that run large volumes of automated text tasks through paid AI APIs.
Key Takeaway
Claude Haiku 5.5 brings Anthropic's entry-level API price down to $0.10 per million input tokens, a cut of up to 90 percent that matches OpenAI's cheapest current tier.
Questions and Answers
When was Claude Haiku 5.5 released?
Anthropic released it on October 7, 2026.
How much does Claude Haiku 5.5 cost?
For prompts up to 100,000 tokens, $0.10 per million input tokens and $0.50 per million output tokens. Above that length, $0.50 and $2.50 respectively.
Is the saving really 90 percent?
On the short-prompt list price, yes, compared with Haiku 4.5. Anthropic's own estimate of the average saving in real use is about 75 percent, because the model's new tokenizer changes token counts.
Who benefits most from the cut?
Companies that send very large numbers of short requests, such as customer-support, document-processing and data-extraction tools, will see the biggest drop in costs.
Disclaimer: Prepared by the Peepals newsroom from publicly available sources with AI assistance. Information is accurate to the best of our knowledge at the time of publication and may change. Images may be representative. Not professional advice. Report an error via our contact page.
Sourced and fact-checked by the Peepals Global Editorial Team
Reported, fact-checked and published by the Peepals Global Editorial Team.
