Claude Haiku 5.5
- Provider
- Anthropic
- Status
- Current
- Context
- 1,000,000 tok
- Price
- $0.1 / $0.5 /MTok
- Knowledge
- 2026-06
Claude Haiku 5.5 is Anthropic’s smallest and fastest current model, released on 7 October 2026 as claude-haiku-5-5 on the Claude API, Amazon Bedrock, Google Cloud and Microsoft Foundry (Anthropic). It succeeds Claude Haiku 4.5 at $0.10/$0.50 per million tokens for prompts up to 100,000 tokens, which Anthropic puts at 90% below Haiku 4.5’s $1/$5. Longer prompts cost $0.50/$2.50.
Artificial Analysis scored it 43 on its Intelligence Index at max effort, at $0.21 per completed task. That is one point under Kimi K3 (44), two above Gemini 3.8 Flash (41) and 13 under Claude Sonnet 5.5 (56). Anthropic published no SWE-bench figure for this release.
Quick specs
| Provider | Anthropic |
| Released | 7 October 2026 (Claude API, Bedrock, Google Cloud, Microsoft Foundry) |
| API model ID | claude-haiku-5-5 |
| Context window | 1,000,000 tokens |
| Max output | 128,000 tokens (300K via the Batch API beta header) |
| Knowledge cutoff | June 2026 |
| Input price | $0.10 / MTok up to 100,000-token prompts; $0.50 above |
| Output price | $0.50 / MTok up to 100,000-token prompts; $2.50 above |
| Cache reads / writes | $0.01 / $0.125 per MTok ($0.05 / $0.625 above 100,000 tokens) |
| Default effort | Medium on the Claude API |
| Modalities | Text and image in; text out |
| AA Intelligence Index | 43 at max effort; 41 xhigh, 38 high, 34 medium, 29 low |
| Cost per AA index task | $0.21 at max effort; $0.08 at high |
| Limitations | No SWE-bench figure; higher prices on prompts over 100,000 tokens; default effort scores nine points under max |
What’s new in Haiku 5.5
Anthropic’s model documentation describes Haiku 5.5 as the model “for high-volume, latency-sensitive tasks such as classification, extraction, and routing”, and its launch post calls it the company’s fastest model to date. Three things changed from Haiku 4.5:
The price. $0.10/$0.50 per million tokens for prompts up to 100,000 tokens, against $1/$5. It is the only current Claude model priced by prompt length: a prompt over 100,000 tokens pays $0.50/$2.50, which is still half Haiku 4.5’s rate.
An effort setting. Anthropic says Haiku 5.5 is its first Haiku-class model with an adjustable effort setting. The Claude API default is medium.
The size of the model’s working room. The context window goes to 1,000,000 tokens and the maximum output to 128,000, the same as Sonnet 5.5 and Opus 5.5, with a June 2026 knowledge cutoff.
Benchmark performance
Anthropic’s launch table
| Benchmark | Claude Haiku 5.5 | Claude Haiku 4.5 | GPT-6 Luna | Claude Sonnet 5.5 |
|---|---|---|---|---|
| GDPval-AA v2.1 | 1620 Elo | 735 | 1437 | 1840 |
| AA-Briefcase v1.1 | 1578 | 614 | 1336 | 1824 |
| OSWorld 2.1 | 72.4% | 15.7% | 48.9% | 83.9% |
| Humanity’s Last Exam (no tools) | 45.9% | 10.2% | — | 56.9% |
| Humanity’s Last Exam (with tools) | 57.4% | 18.7% | — | 64.5% |
| Terminal-Bench 4.0 | 39.2% | 0.0% | 16.4% | 70.6% |
| FrontierCode 1.1 | 46.4% | — | 42.4% | 52.1% (xhigh) |
| Chartography (no tools) | 46.4% | 6.4% | 29.1% | 61.6% |
These are vendor figures. Haiku 5.5 is ahead of GPT-6 Luna on all six rows where Anthropic prints a score for both, and behind Sonnet 5.5 on every row.
The independent read (Artificial Analysis)
| Measure | Claude Haiku 5.5 | Context |
|---|---|---|
| Intelligence Index | 43 (max effort) | Kimi K3 44; Gemini 3.8 Flash 41; Claude Sonnet 5.5 56 |
| Index by effort | 29 low, 34 medium, 38 high, 41 xhigh, 43 max | API default is medium |
| Cost per index task | $0.21 (max); $0.08 (high) | Claude Sonnet 5.5: $7.60 |
| Output speed | ~242 tokens/s (max); ~160 (high) | Claude Sonnet 5.5: ~142 |
Source: Artificial Analysis’s model page and its per-effort results.
Safeguards
Anthropic says Haiku 5.5’s cybersecurity safeguards are more restrictive than Haiku 4.5’s and somewhat less restrictive than those on its other recent models. Its biology safeguards match those on Sonnet 5, Sonnet 5.5 and Opus 5.
Pricing breakdown
| Haiku 5.5 (prompts to 100K) | Haiku 5.5 (prompts over 100K) | Claude Haiku 4.5 | |
|---|---|---|---|
| Input | $0.10 | $0.50 | $1.00 |
| Output | $0.50 | $2.50 | $5.00 |
| Cache reads | $0.01 | $0.05 | $0.10 |
| 5-minute cache writes | $0.125 | $0.625 | $1.25 |
| Batch (in / out) | $0.05 / $0.25 | $0.25 / $1.25 | $0.50 / $2.50 |
Prices per million tokens, from Anthropic’s pricing documentation. Every other current Claude model charges the same rate across its full context window.
How Haiku 5.5 compares
vs Claude Haiku 4.5
A tenth of the token price for prompts up to 100,000 tokens and half above, a 1,000,000-token context window, a June 2026 knowledge cutoff and an effort setting. On Anthropic’s table Terminal-Bench 4.0 goes from 0.0% to 39.2% and OSWorld 2.1 from 15.7% to 72.4%. Haiku 4.5 remains available as a legacy model on the Claude API.
vs GPT-6 Luna
The same $0.10/$0.50 token price for prompts up to 100,000 tokens; GPT-6 Luna keeps that price at any prompt length. Haiku 5.5 leads on every row of Anthropic’s table where both have a score. Artificial Analysis measures Haiku 5.5 at 43 on its index for $0.21 per completed task at max effort.
vs Claude Sonnet 5.5
Sonnet 5.5 scores 13 points higher on the index (56 to 43) at 20 times the token price ($2/$10), and cost $7.60 per completed index task against $0.21. Haiku 5.5 generates output faster: about 242 tokens per second to about 142.
Known limitations
No SWE-bench figure. Anthropic published no SWE-bench Pro or Verified result for this release.
Prompt-length pricing. Prompts over 100,000 tokens cost five times the headline rate.
The default is not the headline. The API default is medium effort, where Artificial Analysis scores the model 34, nine points under its max-effort score.
It trails the larger models. 13 points under Sonnet 5.5 on the index, and behind it on every row of Anthropic’s own table.
Frequently asked questions
When was Claude Haiku 5.5 released?
7 October 2026, as claude-haiku-5-5 on the Claude API, Amazon Bedrock, Google Cloud and Microsoft Foundry.
How much does Claude Haiku 5.5 cost?
$0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens, and $0.50 and $2.50 for longer prompts. Cache reads are $0.01 per million ($0.05 on longer prompts) and batch requests are half price. Claude Haiku 4.5 costs $1 and $5.
Is Claude Haiku 5.5 better than Claude Haiku 4.5?
On Anthropic’s published results, yes on every benchmark where both have a score, including OSWorld 2.1 (72.4% to 15.7%) and GDPval-AA (1620 to 735). It also has a larger context window, a newer knowledge cutoff and a lower price.
What is Claude Haiku 5.5’s context window?
1,000,000 tokens, with text and image input and text output, a 128,000-token maximum output (300,000 through the Batch API beta) and a June 2026 knowledge cutoff.
How does Claude Haiku 5.5 score on independent benchmarks?
Artificial Analysis scores it 43 on its Intelligence Index at max effort, 38 at high and 34 at medium, the API default. A completed index task cost $0.21 at max effort.