Anthropic’s cheapest model now costs $0.10 per million input tokens — identical to OpenAI’s GPT-6 Luna. Here is what is confirmed, what the long-context rule does to the price, and who should switch.
Verified against Anthropic’s launch post and pricing documentation on 21 October 2026.
Claude Haiku 5.5, released 7 October 2026, costs $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens — 90% less than Claude Haiku 4.5 ($1 and $5). It is available on the Claude Platform, AWS, Google Cloud and Azure as claude-haiku-5-5.
The catch: above 100,000 tokens the rates rise to $0.50 input and $2.50 output, and Haiku 5.5 is the stated exception to the flat-price 1M-token window on Claude 4.6-and-later models. At short context it ties OpenAI’s GPT-6 Luna to the cent.
| Item | Confirmed by Anthropic |
|---|---|
| Released | 7 October 2026 |
| API model ID | claude-haiku-5-5 |
| Available on | Claude Platform, Amazon Web Services, Google Cloud, Microsoft Azure |
| Positioning | “Fastest model to date”; high-volume, cost-sensitive work and subagents |
| New | First Haiku-class model with an adjustable effort setting |
| Context window | Not stated in the launch post |
Per million tokens, from Anthropic’s pricing documentation:
| Rate | Prompt up to 100k tokens | Prompt over 100k tokens | Haiku 4.5 |
|---|---|---|---|
| Input | $0.10 | $0.50 | $1.00 |
| Output | $0.50 | $2.50 | $5.00 |
| Cache read | $0.01 | $0.05 | $0.10 |
| Cache write (5 min) | $0.125 | $0.625 | $1.25 |
| Batch input / output | $0.05 / $0.25 | $0.25 / $1.25 | $0.50 / $2.50 |
The headline: for prompts up to 100,000 tokens, Haiku 5.5 is a tenth of the price of Haiku 4.5. The catch: past 100,000 tokens every rate is multiplied by five, and Haiku 5.5 is the stated exception to Anthropic’s rule that Claude 4.6-and-later models include the 1M-token window at one flat price. If your workload is short prompts — classification, routing, extraction, short summaries — you get the full discount. If you feed it long documents, price the long tier, not the headline.
Anthropic describes Haiku 5.5 as built for high-volume, cost-sensitive tasks: summaries, compaction, database queries and classification, and as a subagent that does the cheap, repetitive work while a larger model plans. That subagent pattern is where a model at this price changes the economics — a large model decides what to do once, and a swarm of Haiku calls does it thousands of times.
It is also the first Haiku with an adjustable effort setting, which lets you trade a little cost and latency for more reasoning on harder inputs without moving up a tier.
These are Anthropic’s own numbers from its launch post. We have not reproduced them, and vendor-reported benchmarks are best read as a reason to test, not a result.
| Benchmark | Haiku 5.5 (Anthropic-reported) |
|---|---|
| Humanity’s Last Exam | 45.9% without tools, 57.4% with tools |
| OSWorld 2.1 (offline subset) | 72.4% |
| Terminal-Bench 4.0 | 39.2% |
Whatever you choose, run 30 to 50 real examples from your own workload through each candidate several times and record quality, latency and total cost together. That afternoon of testing settles the question better than any benchmark table, including the ones above.
Haiku 5.5 makes Anthropic’s budget tier a tenth of what it was and ties OpenAI’s on price. For short-prompt, high-volume work, it is worth testing this week.
Price is now a tie at the budget tier, so the deciding factors are accuracy on your task, latency, and how long your prompts really are. Test, do not assume.
Not every AI decision needs a spreadsheet. XAIWebsites generates your complete business website — copy, layout, mobile version and a downloadable site (managed hosting optional) — from a short questionnaire. $199 one-time, and you preview it before you pay.
Anthropic released Claude Haiku 5.5 on 7 October 2026. The API model ID is claude-haiku-5-5, and it is available on the Claude Platform, Amazon Web Services, Google Cloud and Microsoft Azure.
For prompts up to 100,000 tokens: $0.10 per million input tokens and $0.50 per million output tokens, with cache reads at $0.01 and 5-minute cache writes at $0.125. For prompts over 100,000 tokens: $0.50 input and $2.50 output. Batch processing is half of standard rates.
Haiku 4.5 costs $1 input and $5 output. For prompts up to 100,000 tokens, Haiku 5.5 is 90% cheaper on both. Above 100,000 tokens it is 50% cheaper. (Anthropic's launch copy also uses a 75% figure; the 90% and 50% numbers are simply the price ratios from its own table.)
No — they are priced identically at $0.10 input and $0.50 output per million tokens, matching on cached input too ($0.01). The difference is the long-context rule: Haiku 5.5 rises to $0.50 / $2.50 above 100,000 tokens. Since price is a tie for short prompts, choose on accuracy and latency for your own task.
Anthropic positions it for high-volume, cost-sensitive work: summaries, compaction, database queries and classification, and as a subagent alongside larger models. It is also the first Haiku-class model with an adjustable effort setting.
Not at a flat price. Anthropic's pricing page says Claude 4.6 and later models, with Haiku 5.5 as the stated exception, include the full 1M-token window at standard pricing. Haiku 5.5 instead has a two-tier price that steps up after 100,000 tokens. Anthropic's launch post does not state a context-window size.
If your prompts are under 100,000 tokens, the price cut alone makes it worth testing: run 30 to 50 real examples through both and compare quality, latency and cost. Do not switch blind — price falling by a factor of ten says nothing about whether accuracy on your task held.
Figures on this page come from Anthropic’s own pages, read on 21 October 2026. Pricing changes without much notice — confirm before budgeting.
Published 21 October 2026. Claude Haiku 5.5 was released on 7 October 2026. Prices change — check Anthropic’s own pages before budgeting. See our methodology.