Claude Haiku 4.5 vs GPT-4o mini: which is better?

Claude Haiku 4.5 leads on combined benchmark scores (79.8 vs 79.8) across general knowledge, coding, and math, as of 2026-07-27.

Which is cheaper, Claude Haiku 4.5 or GPT-4o mini?

GPT-4o mini is cheaper on a typical API workload, at roughly 7.5x lower cost than the alternative at published per-token rates.

Claude Haiku 4.5 vs GPT-4o mini

Claude Haiku 4.5 leads on benchmarks; GPT-4o mini is cheaper to run, about 7.5× less.

Updated 2026-07-27

Claude Haiku 4.5

Anthropic · Closed

79.8

Combined index

Leads

GPT-4o mini

OpenAI · Closed

79.8

Combined index

Benchmarks

General knowledge (MMLU)

80.1

Coding

79.8

87.2

Math

79.4

70.2

API pricing

$1.0 Input / 1M $0.15

$5.0 Output / 1M $0.6

$45.0 Monthly, est. $6.0

Monthly estimate assumes 20M input + 5M output tokens. Prices marked ~ are representative hosted rates for open-weight models and vary by provider. Run your own numbers in the cost calculator.

Claude Haiku 4.5 costs → GPT-4o mini costs → Full leaderboard → All comparisons →