Claude Haiku 5.5 replaced Haiku 4.5 on Oct 7, 2026 at a tenth of the price and five times the context window. Haiku 4.5 is now a legacy model but still available.
Haiku 5.5 is 90% cheaper than Haiku 4.5 on input, output, cache hits and cache writes for prompts up to 100K tokens. Even its long-prompt rate ($0.50/$2.50 per 1M) is half of Haiku 4.5's standard price — and Haiku 4.5 can't go past 200K tokens at all.
| Claude Haiku 5.5 | Claude Haiku 4.5 | |
|---|---|---|
| Provider | Anthropic | Anthropic |
| Tier | Budget | Budget |
| Input / 1M tokens | $0.10 | $1.00 |
| Output / 1M tokens | $0.50 | $5.00 |
| Cache hit / 1M | $0.01 | $0.10 |
| Cache write / 1M | $0.125 | $1.25 |
| Context window | 1M | 200K |
| Notes | Released Oct 7, 2026; 90% cheaper than Haiku 4.5 with a 1M context window (for prompts ≤100K tokens); >100K tokens: $0.50/$2.50, cache read $0.05, cache write $0.625 | Fast, near Sonnet 4 quality; now a legacy model, superseded by Haiku 5.5 (Oct 7, 2026) but still available |
Claude Haiku 5.5 is 10× cheaper on input tokens ($0.10 vs $1.00 per 1M). Claude Haiku 5.5 is 10× cheaper on output tokens ($0.50 vs $5.00 per 1M).
| Workload | Claude Haiku 5.5 | Claude Haiku 4.5 | Difference |
|---|---|---|---|
| Support chatbot 10,000 conversations × (2,000 input + 400 output tokens) | $4.00 | $40.00 | Claude Haiku 5.5 saves $36.00 |
| Document Q&A 1,000 questions × (30,000 input + 800 output tokens) | $3.40 | $34.00 | Claude Haiku 5.5 saves $30.60 |
| Coding agent session 4M input tokens (90% cache hits) + 150,000 output tokens | $0.15 | $1.51 | Claude Haiku 5.5 saves $1.36 |
Standard rates per 1M tokens from data on this site. The coding-agent row assumes 90% of input is read from the prompt cache (full input price where a model has no cache rate). Cache-write fees, batch discounts and long-context surcharges listed in the notes are not included. Updated November 2, 2026.
you are building anything new or can re-test prompts — it is cheaper at every prompt length and handles up to 1M tokens.
you have a frozen production integration validated on Haiku 4.5 and must keep its exact behaviour until you can re-run evals.
Below 100K tokens, Haiku 5.5 is $0.10/$0.50 against Haiku 4.5's $1.00/$5.00. From 100K up to 4.5's 200K limit, Haiku 5.5 bills $0.50/$2.50 — still half. Beyond 200K, only Haiku 5.5 can take the request.
Haiku 5.5 cache hits are $0.01 and writes $0.125 per 1M, versus $0.10 and $1.25 on 4.5. Anthropic counts cached tokens toward Haiku 5.5's 100K threshold, so a cached 150K prompt is billed at the higher tier — still cheaper than 4.5.
Legacy models tend to be retired eventually, so treat Haiku 4.5 as a temporary fallback. Run your regression set on 5.5, compare output lengths, and keep prompts under 100K tokens where possible to get the full 90% saving. Both models get Anthropic's 50% Batch API discount.
90% on every rate for prompts up to 100K tokens: $0.10/$0.50 vs $1.00/$5.00 per 1M.
Yes. It is now a legacy model, superseded by Haiku 5.5, but still offered on the API.
Haiku 5.5 lists 1M tokens and Haiku 4.5 200K.
How we choose: prices come from each vendor's official pricing page and are re-checked every week. We take no affiliate fees, so rankings aren't for sale.