Anthropic releases the AI model, Claude Haiku 5.5, optimized for speed, lower cost, coding, and large-volume workloads. On Oct. 7, 2026, Anthropic released its small model, Haiku 5.5, which it says is the most affordable and fastest small model yet. It’s meant for use cases including summaries, database lookups, classification, customer support, browser, and AI agent work. According to the company, Haiku 5.5 is 75 percent cheaper on average than the prior model, Haiku 4.5. Haiku 5.5 is available now on the Claude Platform, Amazon Web Services, Google Cloud, and Microsoft Azure, with the model ID claude-haiku-5-5.
Claude Haiku 5.5 Concentrates on Speed and High-Volume Tasks
Anthropic also uses Claude Haiku 5.5 to showcase workloads that require speed and cost efficiency rather than just capability. This model is suited to quick, repetitive tasks that can become expensive at scale. It’s designed for summarization, context reduction, database queries, and classification queries. Anthropic also describes Haiku 5.5 as an effective subagent to larger models like Claude Sonnet 5.5 and Claude Opus 5.5 for coding-oriented workflows, and said it is the company’s fastest model for use in real-time customer service and internet browsing use cases.
Anthropic Reports Major Performance Gains
Haiku 5.5 delivers significant improvements over Haiku 4.5 across several evaluations. Anthropic reports the following results:
Claude Haiku 5.5 also delivers significant improvements over Haiku 4.5 across several evaluations. Anthropic reports a 72.4% score on the OSWorld 2.1 offline subset for computer-use tasks, compared with 15.7% for Haiku 4.5. The model also scores 39.2% on Terminal-Bench 4.0 for agentic coding and 46.4% on FrontierCode 1.1. On the knowledge-work benchmark GDPval-AA v2.1, Haiku 5.5 records a score of 1,620, compared with 735 for Haiku 4.5.
| Benchmark | Haiku 5.5 | Haiku 4.5 | GPT-6 Luna |
| Knowledge work GDPval-AA v2.1 | 1620 | 735 | 1437 |
| Knowledge work AA-Briefcase v1.1 | 1578 | 614 | 1336 |
| Computer use OSWorld 2.1 | 72.4% | 15.7% | 48.9% |
| Multidisciplinary reasoning Humanity’s Last Exam | 45.9% | 10.2% | — |
| Agentic coding Terminal-Bench 4.0 | 39.2% | 0.0% | 16.4% |
| Agentic coding FrontierCode 1.1 | 46.4% | — | 42.4% |
| Visual reasoning Chartography | 46.4% | 6.4% | 29.1% |
These results are company-reported benchmarks, so independent testing will provide additional context as developers use the model.
Claude Haiku 5.5 Adds Adjustable Effort Settings
Anthropic introduces an adjustable effort setting with Haiku 5.5 for the first time in its Haiku model line. This gives developers more control over the balance between cost and intelligence when running workloads. The setting allows users to choose how much effort the model applies to a task. Lower effort can help reduce costs for simpler requests, while higher effort can improve performance when a task requires more reasoning.
Anthropic says this approach helps developers select the right balance based on the needs of each application. Early customers also report improvements in practical workloads. Asana says Haiku 5.5 reduces task-completion latency by more than 30% in its testing, while HubSpot reports a 92.8% average score across its CRM evaluation suite. Box says the model scores 11 points higher than Haiku 4.5 at about half the latency in its early testing.
Anthropic Cuts Claude Haiku 5.5 Pricing
Price is one of the main changes in the new model. Anthropic sets Haiku 5.5 at $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens. For requests above 100,000 tokens, the prices increase to $0.50 for input and $2.50 for output tokens. Anthropic says Haiku 5.5 is priced 90% lower than Haiku 4.5 for requests up to 100,000 tokens and 50% lower for larger requests. Around 90% of requests to Haiku 4.5 fell into the lower-token category, according to the company.
Haiku 5.5 Launches With Wider Availability
Anthropic makes Claude Haiku 5.5 available immediately across its Claude Platform, Amazon Web Services, Google Cloud, and Microsoft Azure. Developers can access the model through the claude-haiku-5-5 model ID on the Claude Platform.
Anthropic also lowers the cache-read price for Claude Sonnet 5.5 by 50%, which the company says reduces its cost on most agentic tasks by around 20%. The company is also introducing monthly API credits for Max and Team subscribers and adding beta support for computer use and browser use in its Claude Python and TypeScript SDKs.
Haiku 5.5 is positioned for narrowly scoped, high-volume work where previous Claude models could become too expensive. Larger models such as Sonnet 5.5 and Opus 5.5 remain better suited to complex agentic coding, while Haiku 5.5 focuses on making everyday AI workloads faster and more affordable.
