Cafiyn Pulse
Anthropic · Verified 2026-09-05

Claude Haiku 4.5 pricing, and what it is actually for.

Anthropic's budget tier at $1 and $5 per million tokens. The model most high-volume traffic should be routed to.

Claude Haiku 4.5
Anthropic · standard tier
Verify against Anthropic's pricing page ↗
$1.00
per 1M input tokens
$5.00
per 1M output tokens
A typical 1,200-in / 400-out token request costs about $0.0032, or roughly $3.20 per 1,000 requests.

Haiku 4.5 runs $1 per million input tokens and $5 per million output, half the price of Sonnet 5 on both legs. For classification, extraction, routing, short summaries and the other high-volume low-complexity work that makes up most production traffic, the quality difference against Sonnet is usually invisible to the user and the cost difference is not.

The mistake teams make is treating model choice as a single decision rather than a per-call-site one. A support product might route intent classification to Haiku, answer generation to Sonnet, and only escalate to Opus on a low-confidence retry. Splitting by call site rather than by product is where the real savings are, and it is why the calculator above lets you model a mix rather than one model.

Best for
  • Classification, extraction, routing and other high-volume structured tasks
  • The default tier in a routed system, with escalation on low confidence
  • Latency-sensitive paths where the user is waiting on the response
Head to head

Claude Haiku 4.5 vs the alternatives.

ModelInput / 1MOutput / 1MSample request
Claude Haiku 4.5$1.00$5.00$0.0032
Gemini 3.8 Flash$0.75$3.75$0.0024
DeepSeek V4 Pro$0.66$1.98$0.0016
Sample request assumes 1,200 input and 400 output tokens, a typical chat exchange. Your actual ratio will differ, run your own numbers in the calculator below.
FAQ

Common questions.

How much cheaper is Haiku 4.5 than Sonnet 5?

Haiku runs $1/million input and $5/million output tokens, versus Sonnet 5 at $2/million input and $10/million output, so half the price on both legs.

Is Claude Haiku 4.5 cheaper than Gemini 3.8 Flash?

No. Gemini 3.8 Flash is $0.75/million input and $3.75/million output against Haiku at $1 and $5, so Flash is cheaper on both. Note that the Flash rate is introductory and Google has published an end date of 31 December 2026, so the comparison changes in January.

More models

Browse other pricing pages.

Open the tool.

Live math against Anthropic's current pricing plus every other provider in the roster.

Compare Claude Haiku 4.5 on my usage