Compare

GPT-5 Mini vs GPT-5 Turbo

OpenAI vs OpenAI. Specs, benchmarks, and real per-task cost — all in one page.

Verdict

GPT-5 Turbo leads on LMSYS ELO (1398 vs 1288). GPT-5 Mini is ~12.0× cheaper on a 3:1 input:output blend.

OpenAI

GPT-5’s smaller, cheaper sibling. Optimized for chat and lightweight agent tasks.

Very low latencyCheap enough for mass consumer apps

OpenAI

OpenAI’s flagship unified model. Handles text, vision, and audio natively. The generalist benchmark champion.

Native multimodalStrong math and scienceHuge ecosystem

Pricing

Input / 1M$0.50$6.00

Output / 1M$2.00$24.00

Context256K400K

Max output16K32K

LicenseProprietaryProprietary

Released2026-01-212026-01-21

Benchmarks

LMSYS ELO1288.01398.0

MMLU Pro78.591.2

HumanEval85.593.0

SWE-Bench—65.8

MATH79.891.5

GPQA—68.1

MMMU—78.8

IFEval85.488.9

Per-task cost

Summarize a 1-hour meeting transcript$0.0085$0.102

Review a 500-line pull request$0.0080$0.096

Answer a customer support ticket$0.0031$0.037

Extract structured data from a resume$0.0045$0.054

Debug a stack trace with context$0.0070$0.084

Per-call cost using published token counts for each task. Real-world prompts vary.

Add up to four models, tweak tasks live.