Claude Haiku 5.5 is on Complete, built for high volume work
New models on Complete
Claude Haiku 5.5 is Anthropic's new small model, released on October 7. It is built for the steps that run thousands of times a day: routing requests, classifying tickets, pulling fields out of documents, summarizing long threads and drafting customer replies. It is available for enterprise teams on Complete.
At its highest effort setting it scores 43 on the Artificial Analysis Intelligence Index, 2nd of 182 models in its price tier on Artificial Analysis, for $0.21 per task. Claude Opus 5.5 scores 58 for $5.98.
Claude Haiku 5.5 | |
|---|---|
Independent score (max effort) | 43 |
Cost per task (max effort) | $0.21 |
Speed | 240 tokens per second |
List price per million tokens in / out* | $0.10 / $0.50 |
Context window | 1 million tokens |
Input | Text and images |
* Providers price per million tokens, which are fragments of words. This price applies to requests under 100,000 tokens.
Above Sonnet 5.5 at medium effort, for a third of the cost
The Artificial Analysis Intelligence Index is an independent rating built from 10 tests covering agent work, coding, general knowledge and science. Cost per task is what Artificial Analysis spent running each model through that suite, divided by the number of tasks.
Haiku 5.5 at max effort outscores Claude Sonnet 5.5 on its medium setting, which scores 41 for $0.59, at about a third of the cost. It also beats GLM-5.3-Flash (42 for $0.25) and DeepSeek V4.1 Flash (39 for $0.27).
It is not the cheapest route to a score in the 40s. MiMo-V2.6-Pro, an open model from Xiaomi, scores 46 for $0.13 per task on the same index. Haiku 5.5 is the lowest cost route to a 40 plus score from Anthropic, with distribution through the major enterprise clouds.
Effort decides the bill
Haiku 5.5 adds an effort setting to the Haiku line. Lower settings cost less and score lower:
Effort setting | Score | Cost per task |
|---|---|---|
Max | 43 | $0.21 |
Extra high | 41 | $0.12 |
High | 38 | $0.08 |
Medium (default) | 34 | $0.05 |
Low | 29 | $0.02 |
Extra high gives up two points against max for a little over half the cost. Medium, the default, runs a task for five cents.
At max effort the model writes a lot before it answers. Artificial Analysis counted 440 million output tokens across its test suite, against a median of 100 million for comparable models. On high volume work, the effort setting is the biggest lever on cost.
Fast enough for live work
Artificial Analysis measures Haiku 5.5 at 240 tokens per second on Anthropic's API, tenth of 182 models and more than twice the median of 110. Anthropic calls it its fastest model at standard speed and points it at live customer service and browser automation, where response time shows up in front of a customer.
Customers quoted by Anthropic report:
Asana: over 30% lower task completion latency.
Box: 11 points higher than Haiku 4.5 at about half the latency.
AlphaSense: 0.84 against 0.76 for Haiku 4.5 on 400 document queries, on a feature that handles about 8 million calls a week.
HubSpot: 92.8% on its internal CRM test suite, averaged over three runs, its best result yet.
Pricing and what changes from Haiku 4.5
Per million tokens | Haiku 5.5, under 100K | Haiku 5.5, over 100K | 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 |
Anthropic puts the average saving against Haiku 4.5 at about 75%, after accounting for a new tokenizer. Batch jobs cost half.
Two changes matter for anyone moving from Haiku 4.5:
The new tokenizer counts the same text as roughly 30% more tokens, per Anthropic's documentation.
The temperature, top_p and top_k settings now return an error. Workflows that set them need updating.
Anthropic lists Haiku 4.5 as active, with retirement not sooner than October 15, 2026.
Anthropic still recommends Claude Sonnet 5.5 and Claude Opus 5.5 for complex coding work. Haiku 5.5 fits the narrower, repeated steps inside larger AI workflows, including running as a subtask worker alongside those models.
Run it on your own workflows
Claude Haiku 5.5 is available for enterprise teams on Complete. Book a demo to run it against your own high volume workflows.
