← Model indexMODEL PROFILE / claude-haiku-4-5

Anthropic / stable

Claude Haiku 4.5

Claude's fast tier, offering near-frontier general capability with lower latency.

Context200K
ModalityText · Image
AccessAPI · Cloud
ReasoningExtended

Where it fits.

Useful for responsive, high-volume product surfaces and as a routing or preprocessing layer before larger Claude models. Haiku 4.5 is Anthropic's fastest near-frontier tier with a 200K context and up to 64K output. Manual extended thinking is available but off by default.

02

Core capabilities

  • 200K context and up to 64K output
  • Fast text and image understanding
  • Optional manual extended thinking
  • Tool use for responsive and high-volume applications
03

Best-fit workflows

  • Responsive support and Q&A
  • Batch text processing
  • Agent subtasks
04 / OPERATIONS

From evaluation
to production.

Deployment notes

Start efficiency-first for prototypes, support surfaces, preprocessors and subagents. Measure accuracy and tail latency, then route difficult code, judgment and long planning cases to Sonnet or above.

Watch before choosing

Escalate difficult coding, ambiguous decisions and long reasoning chains to Sonnet or above. It does not use the same adaptive-thinking configuration as Sonnet 5. Long, ambiguous agent tasks should have an explicit escalation route instead of relying on a larger thinking budget alone.

05

Family & versions

Specifications and positioning checked against provider documentation.

Open official source↗