modelbenchmark.io

GPT-5

OpenAI · gpt-5

Compare

Original GPT-5 workhorse for reasoning, coding, writing, and tool workflows

Specification

openai listing

Context
400K
Max output
128K
Released
2025-08-07
Knowledge cutoff
2024-09-30
Retires
2026-12-11
Open weights
no
Input
text, image
Output
text

Price

US dollars per million tokens · openai list price

Input
$1.25
Output
$10.00
Cache read
$0.125
Cache write
$1.25

Batch and priority prices

US dollars per million tokens · per listing

PriceListing$/M
Batch inputazure$0.6875
Batch inputazure$0.625
Batch outputazure$5.50
Batch outputazure$5.00
Priority inputazure$2.75
Priority inputazure$2.50
Priority outputazure$22.00
Priority outputazure$20.00

Batch input is 45% below the interactive input price at the same listing.

Quality

13 benchmarks · 3 sources

Composite
73rd
percentile of 344 scored models
Rank
93 / 344
±0.04 sd
Evidence
13 × 3
3 or more benchmarks, from 2 or more sources
Effort range
8.8
points between effort settings
coding94th2/3 bench
reasoning55th4/4 bench
math73rd6/7 bench
knowledge68th1/1 bench

By reasoning effort

every setting placed on the same scale as the leaderboard

SettingCompositeBenchmarks behind it
medium85th9
high81st13
low63rd4
minimal34th5

Every score

one row per source and configuration — nothing averaged away

BenchmarkScoreConfigurationSourceRun
Aider polyglot88.0high · aider diffAider2025-08-23
↳86.7medium · aider diffAider2025-08-25
↳81.3low · aider diffAider2025-08-25
Settings differ by 6.7 points on this benchmark.
MATH level 598.1 ±0.3highEpoch AI2025-10-29
↳97.9 ±0.3mediumEpoch AI2025-08-20
Settings differ by 0.2 points on this benchmark.
FrontierMath-v148.6 ±2.6medium · 2 runsEpoch AI2025-11-13
↳46.2 ±2.8high · 2 runsEpoch AI2025-11-13
Settings differ by 2.4 points on this benchmark.
OTIS Mock AIME 2024-202591.4 ±3.7highEpoch AI2025-10-29
↳87.2 ±3.9mediumEpoch AI2025-08-07
↳46.7 ±7.5minimalEpoch AI2026-07-20
Settings differ by 44.7 points on this benchmark.
GPQA diamond86.2 ±2.1highEpoch AI2025-10-29
↳85.3 ±2.1mediumEpoch AI2025-08-07
↳71.7 ±3.2minimalEpoch AI2026-07-20
Settings differ by 14.5 points on this benchmark.
SWE-Bench verified73.5 ±2.0highEpoch AI2026-02-06
↳71.8default · OpenHandsSWE-bench—
↳71.5 ±2.1mediumEpoch AI2026-02-05
↳65.0medium · mini-SWE-agentSWE-bench—
Sources and settings disagree by 8.5 points. Both figures stand.
SimpleQA Verified50.1 ±1.6highEpoch AI2026-08-27
Chess Puzzles36.0 ±4.8highEpoch AI2026-10-13
↳32.0 ±4.7mediumEpoch AI2026-10-13
↳26.0 ±4.4lowEpoch AI2026-10-13
↳10.0 ±3.0minimalEpoch AI2026-10-13
Settings differ by 26.0 points on this benchmark.
FrontierMath-Tier-4-2025-07-016.2 ±0.0high · 2 runsEpoch AI2025-10-30
↳3.1 ±3.5medium · 2 runsEpoch AI2025-08-07
Settings differ by 3.1 points on this benchmark.
FrontierMath-Tiers-1-3-v255.4 ±2.9highEpoch AI2026-06-10
↳37.2 ±2.9lowEpoch AI2026-08-27
↳18.2 ±2.3minimalEpoch AI2026-08-27
Settings differ by 37.2 points on this benchmark.
Mystery Game Puzzles23.0 ±4.2highEpoch AI2026-08-05
↳16.0 ±3.7lowEpoch AI2026-08-27
↳15.0 ±3.6mediumEpoch AI2026-08-27
↳14.0 ±3.5minimalEpoch AI2026-08-27
Settings differ by 9.0 points on this benchmark.
FrontierMath-Tier-4-v221.9 ±6.5highEpoch AI2026-06-11
EBR-bench12.7highEpoch AI2026-06-29

The leading figure for a benchmark is the median across its configurations, so one heroic high-effort run cannot set the number.

Available from 38 hosts

HostIn $/MOut $/MCache rdCache wrContextOutputRetires
openaivendor$1.25$10.00$0.125—
400K
272K
128K2026-12-11
opencode$1.07$8.50$0.107—400K128K—
poe$1.1364$9.0909$0.1136—400K128K—
302ai$1.25$10.00——400K128K—
abacus$1.25$10.00$0.125—400K128K—
aihubmix$1.25$10.00$0.125—400K128K—
aimlapi$1.25$10.00$0.125—400K128K—
azure
$1.25
$1.375
$10.00
$11.00
$0.125
$0.1375
—
400K
272K
128K2027-02-09
azure-cognitive-services$1.25$10.00$0.125—400K128K—
cloudflare-ai-gateway$1.25$10.00$0.125—128K128K—
databricks$1.25$10.00$0.125$1.25
400K
272K
128K—
edenai$1.25$10.00$0.125—400K128K—
fastrouter$1.25$10.00$0.125—400K128K—
gmi · gmi$1.25$10.00——410K32K—
helicone$1.25$10.00$0.125—400K128K—
impossibl$1.25$10.00$0.125—400K128K—
kilo$1.25$10.00$0.125—400K128K—
llmgateway$1.25$10.00$0.125—400K128K—
llmgateway-providers$1.25$10.00$0.125—400K128K—
lowrouter · auto$1.25$10.00$0.125—400K128K—
merge-gateway$1.25$10.00$0.125—400K128K—
nano-gpt$1.25$10.00$0.125—400K128K—
nearai$1.25$10.00$0.125—400K128K—
neon$1.25$10.00$0.125—400K128K—
oci · oci$1.25$10.00——272K128K—
ofox$1.25$10.00$0.13—400K128K—
openrouter$1.25$10.00$0.125—400K128K—
orcarouter$1.25$10.00$0.125—400K128K—
perplexity$1.25$10.00$0.125————
prime$1.25$10.00——400K128K—
replicate · replicate$1.25$10.00—————
sap-ai-core$1.25$10.00$0.125—400K128K—
vercel$1.25$10.00$0.125—400K128K—
zenmux$1.25$10.00$0.125—400K128K—
cortecs$1.375$10.96$0.156—400K128K—
anyapi————400K128K—
model-oracle-ai————400K128K—
qiniu-ai————400K128K—

Price history

append-only observations · a listing writes a row only when its price moves

DateHostField$/M
2026-09-02azureCache read$0.13
2026-10-20azureCache read$0.125
2026-09-02azure-cognitive-servicesCache read$0.13
2026-10-20azure-cognitive-servicesCache read$0.125
2026-09-02ofoxCache read$0.13
2026-09-19ofoxCache read$0.104
2026-10-17ofoxCache read$0.13
2026-09-02openrouterCache read$0.125
2026-10-10openrouterCache read$0.125
2026-09-02poeCache read$0.11
2026-10-22poeCache read$0.1136
2026-09-02zenmuxCache read$0.12
2026-10-22zenmuxCache read$0.125
2026-09-02ofoxInput$1.25
2026-09-19ofoxInput$1.00
2026-10-17ofoxInput$1.25
2026-09-02openrouterInput$1.25
2026-10-10openrouterInput$1.25
2026-09-02poeInput$1.10
2026-10-22poeInput$1.1364
2026-09-02ofoxOutput$10.00
2026-09-19ofoxOutput$8.00
2026-10-17ofoxOutput$10.00
2026-09-02openrouterOutput$10.00
2026-10-10openrouterOutput$10.00
2026-09-02poeOutput$9.00
2026-10-22poeOutput$9.0909
2026-10-13aimlapiCache read$1.7192
2026-10-17aimlapiCache read$0.1625
2026-10-23aimlapiCache read$0.125
2026-10-13aimlapiInput$1.8912
2026-10-17aimlapiInput$1.625
2026-10-23aimlapiInput$1.25
2026-10-13aimlapiOutput$15.1294
2026-10-17aimlapiOutput$13.00
2026-10-23aimlapiOutput$10.00

Other listings of this model

same model, different key

More from OpenAI

most-hosted first