Models/GPT-5 Codex

MDL

GPT-5 Codex

OpenAI

Verified against Artificial Analysis · Jul 13, 2026

Capability

percentile index

Price /M

$1.25 / $10

in / out

Context

400K

tokens

Avg score

68.7

10 benchmarks

AA Intelligence

36.1

index

Speed

175 tok/s

TTFT 6s

Released

Sep 23, 2025

About

GPT-5 Codex is a proprietary model in the GPT family from OpenAI. Benchmarked on 10 evals, averaging 68.7.

Benchmark scores · 10

avg 68.7 — higher is better

AIMEartificial-analysis · 2026-07

98.7

τ²-benchartificial-analysis · 2026-07

86.8

MMLU-Proartificial-analysis · 2026-07

86.5

LiveCodeBenchartificial-analysis · 2026-07

GPQA Diamondartificial-analysis · 2026-07

83.7

IFBenchartificial-analysis · 2026-07

74.1

AA Long-Context Reasoningartificial-analysis · 2026-07

SciCodeartificial-analysis · 2026-07

40.9

Terminal-Bench Hardartificial-analysis · 2026-07

37.9

Humanity's Last Examartificial-analysis · 2026-07

25.6

Details

PublisherOpenAI

FamilyGPT

Parameters—

License—

WeightsClosed

ReleasedSep 23, 2025

Performance & price

Output speed175 tok/s

Latency (TTFT)6s

Price in / out$1.25 / $10 per 1M

Blended price$3.44 / 1M

Math index98.7

Median across API providers, via Artificial Analysis. Blended = 3:1 input:output. For reasoning models, latency includes thinking time. Methodology →

History

full ledger →

Capability over time

Append-only ledger — every observed change to this model's numbers.

API providers

Artificial Analysis$10/M