Term
Claude Haiku 4.5
Claude Haiku 4.5 is Anthropics fast compact model from October 2025 — it delivers near-Sonnet-4 performance at roughly one third of the cost and is the first Haiku model with reasoning.
Claude Haiku 4.5 — explained in more detail
Anthropic released Claude Haiku 4.5 on October 15, 2025 as the fastest and cheapest model in the Claude 4.5 family. The Haiku line has always stood for low latency and low cost; with version 4.5 it reaches, for the first time, a level of performance within roughly five percentage points of the then best-in-class model. Anthropic frames Haiku 4.5 with the line that “what was recently at the frontier is now cheaper and faster”: the model delivers quality comparable to Sonnet 4 at about one third of the cost and roughly twice the speed.
Key facts
- Release: October 15, 2025, proprietary access via Claude.ai, the Anthropic API, Amazon Bedrock and Google Cloud Vertex AI.
- Pricing: $1 per 1M input tokens, $5 per 1M output tokens (about 25 percent above Haiku 3.5, but a third of Sonnet 4.5 pricing).
- Context window: up to 200,000 tokens, maximum 64,000 output tokens (Haiku 3.5 was capped at 8,192).
- Reasoning: first Haiku model with extended-thinking capability.
- Knowledge cutoff: February 2025.
Example / Practical use
Haiku 4.5 is built for tasks that require high volume and low latency — classification, extraction, chat support, or acting as a fast sub-agent in multi-agent systems where a stronger model such as Sonnet or Opus handles orchestration and Haiku works through the parallel sub-steps. Anthropic additionally trained the model against “agentic laziness”, the tendency to report tasks as finished prematurely.
Delimitation
Haiku 4.5 is the entry tier of the Claude 4.5 line: faster and cheaper than Sonnet 4.5 (mid tier) and Opus (top tier), but with less reasoning depth on the hardest tasks. Compared with the cheapest competitor models such as GPT-4o mini or Gemini Flash Lite, Haiku 4.5 sits higher on price but offers markedly stronger coding and agentic performance. Where maximum quality matters, Sonnet or Opus remain the choice.