Anthropic has launched Claude Haiku 5.5, a new small AI model designed for high-volume, cost-sensitive workloads such as summarization, classification, database queries, customer support and AI-agent tasks. The company says the model costs about 75% less to run than Claude Haiku 4.5 on average while delivering significant improvements across coding, computer use and knowledge work.
Claude Haiku 5.5 became available on October 7, 2026, and is available through Anthropic’s Claude Platform as well as Amazon Web Services, Google Cloud and Microsoft Azure. Developers can access it through the model ID claude-haiku-5-5.
The launch also introduces adjustable effort to the Haiku family for the first time, allowing developers to make a trade-off between model intelligence and cost depending on the task.
Quick Summary
- Anthropic launched Claude Haiku 5.5 on October 7, 2026.
- It is designed for fast, high-volume and cost-sensitive AI workloads.
- Anthropic says it costs around 75% less on average than Haiku 4.5.
- It is the first Haiku model with adjustable effort, allowing users to balance intelligence and cost.
- Anthropic reports major improvements in coding, computer use and knowledge work.
- Haiku 5.5 is designed to work alongside Claude Sonnet 5.5 and Claude Opus 5.5 as a subagent.
- It is available through Anthropic, Amazon Web Services, Google Cloud and Microsoft Azure.
- Anthropic also reduced Sonnet 5.5 cache-read pricing, making agentic workloads cheaper.
What Is New in Claude Haiku 5.5?
Claude Haiku 5.5 is positioned as Anthropic’s fastest and most cost-efficient model for workloads where large frontier models may be unnecessarily expensive.
Anthropic says the model is designed for repetitive, high-volume work, including summaries, context compaction, database queries and classification. It can also operate as a subagent alongside larger Claude models such as Opus 5.5 and Sonnet 5.5, particularly for coding workflows.
The company is also targeting speed-sensitive applications such as live customer support and browser-based tasks.
One of the more significant changes is the introduction of an adjustable effort setting. Developers can control how much effort the model applies to a task, allowing them to prioritize lower costs for straightforward requests or higher intelligence when a task requires additional reasoning.
Claude Haiku 5.5 Pricing
Cost is one of the main changes with the new model.
Anthropic lists Claude Haiku 5.5 at:
| Pricing Tier / Metric | Claude Haiku 5.5 Rate |
|---|---|
| Input tokens (prompts up to 100K) | $0.10 per million |
| Input tokens (prompts over 100K) | $0.50 per million |
| Output tokens (prompts up to 100K) | $0.50 per million |
| Output tokens (prompts over 100K) | $2.50 per million |
| Cache reads (up to 100K) | $0.01 per million |
| Cache reads (over 100K) | $0.05 per million |
| Cache writes (up to 100K) | $0.125 per million |
| Cache writes (over 100K) | $0.625 per million |
Anthropic says Haiku 5.5 is priced 90% lower than Haiku 4.5 for requests up to 100,000 tokens and 50% lower for requests above that threshold. Because about 90% of requests to the previous Haiku model were within the lower token range, Anthropic estimates an average operating-cost reduction of roughly 75%.
The company notes that the calculation also accounts for changes in token usage resulting from Haiku 5.5’s updated tokenizer.
Claude Haiku 5.5 Performance Improvements
Anthropic reports improvements across several areas compared with Haiku 4.5.
On its published evaluations, Haiku 5.5 scores substantially higher than the previous model on knowledge work, computer use, multidisciplinary reasoning and agentic coding.
For example, Anthropic reports a 72.4% score on the offline subset of OSWorld 2.1, compared with 15.7% for Haiku 4.5. On Terminal-Bench 4.0, which evaluates agentic coding, Haiku 5.5 scored 39.2%, compared with 0% for Haiku 4.5.
Anthropic also reports a 46.4% result on the main FrontierCode 1.1 evaluation and a 45.9% score on Humanity’s Last Exam without tools.
These are Anthropic’s reported benchmark results, so they should be viewed in the context of the company’s evaluation methodology rather than treated as universal measures of real-world performance.
Designed for AI Agents and High-Volume Work
The positioning of Claude Haiku 5.5 is particularly relevant to developers building AI agents.
Instead of using a large model for every step of an agent workflow, developers can assign smaller, repetitive tasks to Haiku 5.5 while reserving more capable models for complex reasoning and coding.
Anthropic explicitly describes Haiku 5.5 as suitable for subagent work. Its examples include using the model for short lookups, extracting information from documents and handling repetitive operations while a larger model manages the broader task.
This approach can reduce the cost of agentic systems where thousands or millions of relatively small model calls are required.
The model’s speed also makes it relevant to applications where response latency matters. Anthropic specifically highlights live customer support and browser use as areas where Haiku 5.5’s combination of speed, capability and price can be useful.
Adjustable Effort Adds More Cost Control
The adjustable effort setting gives developers another way to manage the economics of model usage.
A simple classification request may not require the same level of reasoning as a multi-step computer-use task. With Haiku 5.5, developers can adjust the model’s effort depending on the complexity of the request.
That makes the model particularly suited to applications with mixed workloads, where optimizing every request for maximum intelligence could unnecessarily increase operating costs.
Anthropic’s published performance-versus-cost evaluations show how the model behaves at different effort levels, reinforcing the company’s focus on giving developers more control over the cost-performance trade-off.
Safety Improvements
Anthropic also reports safety improvements with Claude Haiku 5.5.
According to the company, the model shows major improvements across almost all of its alignment evaluations compared with Haiku 4.5, including fewer instances of behavior Anthropic classifies as misaligned and lower willingness to cooperate with misuse.
Its cybersecurity safeguards have also been updated. Anthropic says Haiku 5.5 allows a wider range of defensive cybersecurity tasks than the safeguards applied to Sonnet 5.5, while continuing to block penetration testing and other techniques considered more likely to be used by attackers.
These results and safeguards are detailed further in Anthropic’s system-card documentation.
Where Claude Haiku 5.5 Is Available?
Claude Haiku 5.5 is available now across Anthropic’s supported platforms, including Amazon Web Services, Google Cloud and Microsoft Azure.
Developers using Anthropic’s own Claude Platform can access the model using claude-haiku-5-5.
Anthropic is also updating its Python and TypeScript SDKs with beta support for computer use and browser use, capabilities that fit the model’s intended use in fast, cost-sensitive agent workflows.
Anthropic Also Cuts Sonnet 5.5 Cache-Read Costs
The Haiku 5.5 launch includes another pricing change for Anthropic’s model lineup.
The company has reduced the cache-read price for Claude Sonnet 5.5 by 50%, from $0.20 to $0.10 per million tokens. Anthropic estimates that the change reduces Sonnet 5.5’s cost on most agentic tasks by around 20%.
Anthropic is also rolling out monthly API credits for Max and Team subscribers, intended to help users experiment with applications, tools and AI agents built on the Claude Platform.
The combination of a lower-cost Haiku model, cheaper Sonnet cache reads and additional API credits shows Anthropic is focusing not only on model capability but also on the economics of running AI applications at scale.
For developers building high-volume AI systems, the main significance of Claude Haiku 5.5 is therefore its balance between speed, capability and operating cost rather than raw frontier-model performance. Anthropic itself continues to position Sonnet 5.5 and Opus 5.5 as better choices for complex agentic coding tasks, while Haiku 5.5 is aimed at narrower, repetitive and cost-sensitive workloads.
Also Read –
Claude Opus 5.5: Everything You Need to Know About Anthropic’s AI Model
Claude Opus 5.5 Launches With Stronger Coding and Lower Costs
Claude Opus 5.5 Reportedly in Testing Ahead of Launch
