THE FORGE RANKINGS·2026-08-30

GPT-5 (2025-08-07)

OpenAI
27
RANK of 37 ranked
FERROX INDEX45.2 weighted mean of category positions
GRADED back of the field
EVIDENCE11 / 32 benchmarks measured
QUALITYFULL every category rests on independent benchmarks

No measurement in Preference. That weight was redistributed, which is the same as assuming this model would have scored its own average there. It is an assumption, not a measurement.

Compared to what

Each rail is one classification. Every ranked model measured on it is a tick; this model is the orange marker. A hatched rail means this model has no measurement in that classification and its weight was redistributed.

Agents

20of 37 measured
38.2
13.1median 38.963.0

Coding

4of 37 measured
60.4
22.4median 50.070.7

Reasoning

28of 37 measured
42.5
18.1median 62.893.0

Preference

no qualifying measurement
53.1median 71.882.5

Every measurement, and the field behind it

One card per benchmark this model has been measured on. The rail shows where it sits against every other ranked model measured on the same benchmark. Where the evaluator keys on a harness, the harness is named, because the same model scores differently under a different one.

Not measured

21/32
  • LMArena Agent
  • LMArena Text (style-controlled)
  • DeepSWE
  • FrontierCode
  • LMArena WebDev
  • Terminal-Bench 4.0
  • OSWorld
  • Cybench
  • BrowseComp
  • IMOAnswerBench
  • SWE-bench Multilingual
  • AIME
  • CyberGym
  • HMMT
  • MCP-Atlas
  • SWE-bench Pro
  • Tau2-Bench Airline
  • Tau2-Bench Banking
  • Tau2-Bench Retail
  • Tau2-Bench Telecom
  • Tool-Decathlon
A missing benchmark is not a zero and is never scored as one.

Check it yourself

Every number on this page, with who measured it, the interval they published, the harness it was run under, and where to go and read it. Nothing here was measured by Ferrox Labs.

BenchmarkClassPublishedInterval (normalised) Measured byLicenceHarness / effort
APEX Agents agents 18.3% 12.6 to 24.0 APEX CC-BY-4.0 source
METR time horizons agents 2.3h 52.1 to 69.5 METR - Measuring AI Ability to Complete Long Tasks CC-BY-4.0 / medium source
Terminal-Bench 2 agents 35.2% 29.1 to 41.3 Terminal-Bench v2 Leaderboard CC-BY-4.0 Terminus 2 / medium source
Aider Polyglot coding 86.7% not published aider.chat CC-BY-4.0 diff / medium source
ALE-Bench coding 808 not published ALE-Bench (Sakana AI with AtCoder) CC-BY-4.0 / minimal source
SWE-bench Verified (Epoch's own run) coding 71.5% 67.5 to 75.5 Epoch AI CC-BY-4.0 / medium source
ARC-AGI-2 reasoning 1.9% not published ARC Prize CC-BY-4.0 / low source
FrontierMath reasoning 27.2% 22.1 to 32.4 Epoch AI CC-BY-4.0 / medium source
GPQA Diamond (Epoch's own run) reasoning 85.4% 81.2 to 89.5 Epoch AI CC-BY-4.0 / medium source
Humanity's Last Exam reasoning 25.3% 22.0 to 28.6 Humanity’s Last Exam (CAIS / Scale AI) CC-BY-4.0 source
OTIS Mock AIME 2024-2025 reasoning 87.2% 79.6 to 94.8 Epoch AI CC-BY-4.0 / medium source

Configurations rolled up: 12. Rule: median observed configuration per benchmark (lower median, always a real measurement). Harnesses seen: codex-cli, diff, mini-swe-agent, openhands, terminus-2.

Snapshot 2026-08-30-62e4e85f043b, manifest dc46131315046f1e. Grades are positional: position in the measured field, as a percentile of rank among ranked entries, n=37.