MODELS
Claude Haiku 4.5
Anthropic's small, fast Claude — cheap tool-calling and vision with 200K context.
Specs
- Context window
- 200,000
- Max output
- 8,192
- Modalities
- text, image
- Tool use
- ✓
- Vision
- ✓
- Streaming
- ✓
- License
- proprietary
- Released
- 2025-10-01
Pricing
- Input / 1M
- $1.00
- Output / 1M
- $5.00
- Cached input / 1M
- $0.10
Cost estimate
Haiku 4.5 is Anthropic's lightweight Claude tier, aimed at high-volume agent loops, classification, and coding sub-tasks where Sonnet is overkill. It keeps the full 200K context, vision input, and tool use, at $1/$5 per million tokens. Best fit: builders who need Claude-style instruction-following and JSON reliability without Sonnet's price tag.
Editor's verdict
Pick Haiku 4.5 when you're running an agent that calls tools thousands of times a day and Sonnet's bill hurts — its tool-use reliability is the main reason to choose it over GPT-4.1 mini or Gemini 2.5 Flash. Reasoning depth on hard coding or math is clearly behind Sonnet 4.5 and GPT-5 mini, so don't use it as your top-of-funnel planner. Output is also pricey at $5/M compared to Gemini Flash.
Reviews
No reviews yet. Be the first.
Last updated: 2026-04-29