On October 15, 2025, Anthropic released Claude Haiku 4.5. The model ID is claude-haiku-4-5-20251001 (alias claude-haiku-4-5), and Anthropic describes it as "the fastest model with near-frontier intelligence." Pricing is $1 / $5 per million input/output tokens — cost-efficient, as expected from the Haiku tier.
Highlights
- Speed and cost efficiency — Anthropic states that Haiku 4.5 matches the coding capability of the earlier Claude Sonnet 4 at roughly one-third the cost and more than twice the speed.
- Parallel orchestration — Anthropic gives the example of Sonnet 4.5 breaking a complex problem into multi-step plans, then orchestrating a team of multiple Haiku 4.5s to complete subtasks in parallel.
- Extended thinking — Haiku 4.5 supports extended thinking. It does not support adaptive thinking.
- Agentic and real-time work — Its fast response makes it well suited to latency-sensitive tasks such as real-time customer service and chatbots.
Spec summary
- Model ID:
claude-haiku-4-5-20251001(aliasclaude-haiku-4-5) - Pricing:
$1input /$5output per MTok - Context window: 200K tokens
- Max output: 64K tokens (sync Messages API)
- Extended thinking: Yes · Adaptive thinking: No
- Comparative latency: Fastest
- Reliable knowledge cutoff: February 2025
Figures are based on Anthropic's official model docs (platform.claude.com) and the launch announcement (anthropic.com/news/claude-haiku-4-5). For specific benchmark scores, please refer to the official announcement.