Compare
o4 vs GPT-5 Turbo
OpenAI vs OpenAI. Specs, benchmarks, and real per-task cost — all in one page.
Verdict
GPT-5 Turbo is ~2.5× cheaper on a 3:1 input:output blend.
OpenAI
o4Deep-reasoning model. Spends more tokens thinking to crush math, science, and hard coding problems.
State-of-the-art math and scienceExtremely high-ceiling reasoning
OpenAI
GPT-5 TurboOpenAI’s flagship unified model. Handles text, vision, and audio natively. The generalist benchmark champion.
Native multimodalStrong math and scienceHuge ecosystem
Pricing
Input / 1M$15.00$6.00
Output / 1M$60.00$24.00
Context200K400K
Max output100K32K
LicenseProprietaryProprietary
Released2026-02-182026-01-21
Benchmarks
LMSYS ELO—1398.0
MMLU Pro93.891.2
HumanEval92.393.0
SWE-Bench68.065.8
MATH96.291.5
GPQA82.568.1
MMMU—78.8
IFEval—88.9
Per-task cost
Summarize a 1-hour meeting transcript$0.255$0.102
Review a 500-line pull request$0.240$0.096
Answer a customer support ticket$0.093$0.037
Extract structured data from a resume$0.135$0.054
Debug a stack trace with context$0.210$0.084
Per-call cost using published token counts for each task. Real-world prompts vary.