Claude Haiku 4.5 is a large language model developed by Anthropic, released in October 2025 as the fourth-generation entry in the Haiku line, the smallest and fastest tier of the Claude model family. Its release was notable for how far it narrowed the gap with frontier models: on several coding and agentic benchmarks it matched or approached Claude Sonnet 4 results from just months earlier, at roughly one third of the cost and more than twice the speed.
Design goals
The Haiku tier targets workloads where volume, latency and cost dominate the engineering tradeoff:
- Real-time applications: chat assistants, customer support, in-product copilots where sub-second response matters.
- High-volume pipelines: classification, extraction, moderation and translation at scale.
- Sub-agent architectures: multi-agent systems where a frontier model orchestrates fleets of cheaper workers, a pattern Anthropic explicitly promoted with the release, positioning Haiku 4.5 as the worker tier under Sonnet or Opus orchestrators.
Claude Haiku 4.5 supports extended thinking (visible chain-of-thought reasoning), tool use and computer use, capabilities that in prior generations were reserved for the larger tiers.
Reception
The model was received as evidence of rapid cost-performance compression in the industry: capabilities that defined the frontier in early 2025 became commodity-priced by the end of the year. On coding leaderboards it consistently ranked as the strongest model in its price class. On Wikiprompt, prompts targeting Haiku 4.5 skew toward high-volume text workflows: summarization templates, extraction schemas and agent system prompts.