Claude Haiku 5.5 Pricing
Last updated: October 8, 2026
Claude Haiku 5.5 is Anthropic's cheapest small model: $0.10 per million input tokens and $0.50 per million output tokens, or around 75% less to run than Haiku 4.5, per Anthropic's release on October 7, 2026.[1] The model ID is claude-haiku-5-5.
The complete published price sheet is below — input, output, and cache rates, including the two prompt-length tiers Anthropic prices separately. After it: what the 75% headline actually rests on, and what the release still does not say.
Claude Haiku 5.5 price table
| Price per 1M tokens | Haiku 5.5 (prompt ≤100k) | Haiku 5.5 (prompt >100k) | Haiku 4.5 | Sonnet 5.5 |
|---|---|---|---|---|
| Cache reads | $0.01 | $0.05 | $0.10 | $0.10 |
| Cache writes | $0.125 | $0.625 | $1.25 | $2.50 |
| Input tokens | $0.10 | $0.50 | $1.00 | $2.00 |
| Output tokens | $0.50 | $2.50 | $5.00 | $10.00 |
Per-token rates exactly as published in Anthropic's release table.[1] Haiku 5.5 is billed on two prompt-length tiers — up to 100k tokens, and above 100k tokens. Sonnet 5.5's cache-read rate was also cut in half on the same day, from $0.20 to $0.10 per million tokens.
What “around 75% less” is based on
Anthropic's headline for Haiku 5.5 is that it costs around 75% less to run than Haiku 4.5. The footnote beneath that figure splits it by prompt length: 90% cheaper for prompts of up to 100k tokens, and 50% cheaper for prompts above 100k. Anthropic also adds a qualifier — the comparison already factors in Haiku 5.5's updated tokenizer, which produces slightly more tokens per task.
That qualifier is why the headline is a range rather than a single number. Sticker price and cost per task stop being the same thing once the tokenizer changes: if your workload mostly sends short prompts, the 90% footnote is the closer reference point; long-prompt pipelines should plan against the 50% figure. Either way, every published per-token rate for Haiku 5.5 sits below the Haiku 4.5 rate in the same category — input $0.10 vs $1.00, output $0.50 vs $5.00, cache reads $0.01 vs $0.10, and cache writes $0.125 vs $1.25 per million tokens at the ≤100k tier.
To model your own bill, run your traffic through the updated tokenizer and price the resulting token counts against the table above. The release page does not publish a per-task conversion, so this page does not invent one.
What you get at these rates
Anthropic positions Haiku 5.5 for work that is frequent, narrow, and previously hard to justify at frontier prices: summaries, compactions, database queries, classification requests, live customer support, and browser use. It also carries a coding-subagent role alongside Opus 5.5 and Sonnet 5.5 in multi-model setups.[1]
Two capability notes at this price point. It is the first Haiku-tier model with an adjustable effort setting, which trades cost against intelligence per request — Anthropic published that trade-off as a chart only, so per-level scores are marked Not yet confirmed here, not guessed. And it is available on the Claude Platform, Amazon Web Services, Google Cloud, and Microsoft Azure.[1]
What Anthropic has not published
The release page does not state the context window, rate limits, or real-world throughput for Haiku 5.5 — so this page marks them Not yet confirmed rather than filling the gaps with third-party numbers. If any of those decides your architecture, treat them as open questions until the platform documentation catches up.
Frequently asked questions
How much does Claude Haiku 5.5 cost?
For prompts of up to 100k tokens, Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens. Above 100k tokens, the rates are $0.50 input and $2.50 output. Cache reads cost $0.01–$0.05 and cache writes $0.125–$0.625 per million tokens across the same two prompt-length tiers.
Is Claude Haiku 5.5 really 75% cheaper than Haiku 4.5?
That is Anthropic's own headline figure: the release says Haiku 5.5 costs around 75% less to run than Haiku 4.5. The footnote splits it further — 90% cheaper for prompts of up to 100k tokens and 50% cheaper for prompts above 100k — and notes the calculation already includes Haiku 5.5's updated tokenizer.
What does the ≤100k / >100k split mean?
Anthropic prices Haiku 5.5 on two tiers by prompt length. Prompts of up to 100k tokens bill at the lower rates; prompts above 100k tokens bill at the higher rates shown in the table. Both tiers are published on the release page.
Why does the tokenizer change matter for cost?
Anthropic notes that Haiku 5.5 uses an updated tokenizer that produces slightly more tokens per task, and that the 75% headline — along with the 90% / 50% footnote — already accounts for that effect. Practically, compare cost per task on your own traffic rather than the per-token sticker price alone.
Where is Claude Haiku 5.5 available?
Anthropic lists it on the Claude Platform, Amazon Web Services, Google Cloud, and Microsoft Azure. Region-level availability is not stated on the release page.
Sources
- Anthropic — Claude Haiku 5.5 (official release page) Published October 7, 2026.
Every figure on this page comes from Anthropic's official Claude Haiku 5.5 release page, published October 7, 2026. VerdictApi did not use third-party estimates; anything the release does not state is labeled Not yet confirmed.
Last updated: October 8, 2026