Claude Haiku 5.5: Cheapest and Fastest Small Model

Features

• ✓ Pricing: Anthropic's pricing table shows $0.10 per million input tokens, $0.50 output and $0.01 cache reads for prompts up to 100,000 tokens, rising to $0.50 input, $2.50 output and $0.05 cache reads beyond that, and says the cheapest tier is 90% lower than Haiku 4.5
• ✓ Positioning: Anthropic says it is built for high-volume, cost-sensitive tasks (summaries, compactions, database queries, classification), pairs well with Opus 5.5 and Sonnet 5.5 as a subagent on coding work, and suits speed-sensitive use such as live customer support and browser use
• ✓ Benchmarks: Anthropic reports 1620 on GDPval-AA v2.1 (versus 735 for Haiku 4.5), 72.4% on the OSWorld 2.1 offline subset (versus 15.7%), 45.9% on Humanity's Last Exam without tools, 39.2% on Terminal-Bench 4.0 and 46.4% on FrontierCode 1.1 Main
• ✓ Adjustable effort: Anthropic says Haiku 5.5 is the first Haiku-class model to come with an effort setting, letting users decide whether to optimise for cost or intelligence
• ✓ Same-day value changes: Anthropic says it cut the price of Sonnet 5.5's cache reads in half to $0.10 per million tokens starting the same day (reducing Sonnet 5.5's cost on most agentic tasks by around 20%) and introduced a new monthly API credit for Claude Max and Team subscribers

How to Use

1. Call it on the Claude Platform with the model ID claude-haiku-5-5 (also on AWS, Google Cloud and Microsoft Azure)
2. Move high-volume short tasks such as classification, extraction, summaries and routing from bigger models to Haiku 5.5 to cut cost
3. Use Haiku 5.5 as a subagent alongside Opus 5.5 / Sonnet 5.5 for fetching data and small steps inside long chains
4. Tune the effort setting to balance cost and accuracy for your workload