Overview / Description
Claude Haiku 5.5 is an AI developer tool — a small language model from Anthropic aimed at developers building high-volume, cost-sensitive applications. Launched October 7, 2026, it is described as the cheapest, fastest, and most capable small model Anthropic has released, and it carries the model ID claude-haiku-5-5. The concrete draw is price paired with capability: for requests up to 100,000 tokens — covering roughly 90% of prior Haiku usage — it is priced at $0.10 per million input tokens, $0.50 per million output, and $0.01 for cache reads, about a 75% average cost reduction versus Haiku 4.5. An adjustable effort setting lets you dial each request toward lower cost or higher intelligence, which is useful when the same model serves both throwaway classification and harder reasoning. It supports browser use and computer use through updated SDKs and is available on AWS, Google Cloud, and Microsoft Azure. Published benchmarks include 72.4% on OSWorld 2.1 for computer tasks, 39.2% on Terminal-Bench 4.0 for agentic coding, and 45.9% on Humanity's Last Exam. For a BestAIFor reader choosing a model for summaries, classification, live support, browser automation, or coding subagents, it targets the speed-and-cost tier rather than frontier reasoning.
Used For
Summaries, classification, live support, browser and computer automation, and coding subagents at low cost
Pricing
Pros & Cons
Pros
- About 75% average cost cut vs Haiku 4.5: $0.10 input / $0.50 output per million tokens (up to 100K tokens)
- Adjustable effort setting to trade cost against intelligence per request
- Supports browser use and computer use via updated SDKs
- Available on AWS, Google Cloud, and Microsoft Azure
- Published benchmarks: 72.4% OSWorld 2.1, 39.2% Terminal-Bench 4.0, 45.9% Humanity's Last Exam
Cons
- Small model tuned for cost and speed, not frontier-level reasoning
- Lowest advertised pricing applies up to 100,000 tokens; longer contexts may price differently
- Benchmark scores trail larger Claude models on hard reasoning tasks
- Realizing best results often requires tuning the effort setting per workload
Questions & Answers
Alternatives
GPT-5 mini, Gemini Flash, Claude Sonnet
Reviews & Ratings
0 reviews
Sign in to rate and review Claude Haiku 5.5.
Sign in to reviewNo reviews yet. Be the first to review Claude Haiku 5.5!