Code generation: a model at 1/271st the price of the best scorer, at 97.8% of its quality.

claude-haiku-4.5 pricing (anthropic)

the measured answer

claude-haiku-4.5 costs $1.00 per million input tokens and $5.00 per million output tokens. On Potion's live measurements it runs from $1.12 per 1,000 requests, and its strongest measured workload is summarization (quality 0.800).

per-token list price + measured per-request cost from live traffic-shaped suites · method

Per-token prices say little about what a request costs: input/output profiles differ per workload, and reasoning-style models spend hidden tokens. The table below is the number a bill is made of — measured cost per 1,000 requests, with the measured quality it buys, per kind of work this model appears on the live frontier for.

$1.00
per 1M input tokens
$5.00
per 1M output tokens

Measured, per kind of work

kind of workmeasured quality$ / 1K requestsp95 latency
Summarization0.800$1.123051 msfrontier →

a model appears here only for workloads where it sits on the live measured frontier — absence means it was measured and beaten, or not yet measured

the routed alternative

Potion routes each request to the cheapest option measured at your quality bar — this model where it earns the route, something cheaper where it does not. Get an API key.

claude-haiku-4.5 pricing & measured cost per request · Frontier Notes