THE FORGE RANKINGS·2026-08-30

Claude 4.1 Opus

Anthropic
32
RANK of 37 ranked
FERROX INDEX34.9 weighted mean of category positions
GRADEE bottom of the measured field
EVIDENCE11 / 32 benchmarks measured
QUALITYTHIN Correlated evidence in: preference. Those category scores rest on benchmarks from a single family.

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

12of 37 measured
45.3
13.1median 38.963.0

Coding

24of 37 measured
46.3
22.4median 50.070.7

Reasoning

30of 37 measured
39.9
18.1median 62.893.0

Preference

14of 15 measured
63.4
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.

ALE-Bench

32/35
675 best 2177 · GPT-5.6 Sol
ALE-Bench (Sakana AI with AtCoder) CC-BY-4.0 16k

SWE-bench Verified (Epoch's own run)

12/21
73.3% best 83.5% · Claude Opus 4.7
Epoch AI CC-BY-4.0

LMArena Text (style-controlled)

13/20
1448 best 1507 · Claude Fable 5
LMArena CC-BY-4.0

LMArena WebDev

14/16
1389 best 1626 · Claude Fable 5
LMArena CC-BY-4.0

Cybench

2/5
42.0% best 55.0% · Claude Sonnet 4.5
Cybench leaderboard CC-BY-4.0

Not measured

21/32
  • ARC-AGI-2
  • APEX Agents
  • LMArena Agent
  • DeepSWE
  • FrontierCode
  • Terminal-Bench 4.0
  • Aider Polyglot
  • OSWorld
  • 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
Cybench agents 42.0% not published Cybench leaderboard CC-BY-4.0 source
METR time horizons agents 1.9h 50.3 to 66.3 METR - Measuring AI Ability to Complete Long Tasks CC-BY-4.0 / 16k source
Terminal-Bench 2 agents 35.1% 30.2 to 40.0 Terminal-Bench v2 Leaderboard CC-BY-4.0 Mini-SWE-Agent source
ALE-Bench coding 675 not published ALE-Bench (Sakana AI with AtCoder) CC-BY-4.0 / 16k source
SWE-bench Verified (Epoch's own run) coding 73.3% 69.4 to 77.3 Epoch AI CC-BY-4.0 source
LMArena Text (style-controlled) preference 1448 77.8 to 78.7 LMArena CC-BY-4.0 source
LMArena WebDev preference 1389 47.3 to 50.0 LMArena CC-BY-4.0 source
FrontierMath reasoning 5.9% 3.1 to 8.6 Epoch AI CC-BY-4.0 source
GPQA Diamond (Epoch's own run) reasoning 76.8% 70.9 to 82.7 Epoch AI CC-BY-4.0 / 27k source
Humanity's Last Exam reasoning 7.9% 5.8 to 10.0 Humanity’s Last Exam (CAIS / Scale AI) CC-BY-4.0 source
OTIS Mock AIME 2024-2025 reasoning 64.4% 50.3 to 78.6 Epoch AI CC-BY-4.0 / 16k source

Configurations rolled up: 7. Rule: median observed configuration per benchmark (lower median, always a real measurement). Harnesses seen: claude-code, 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.