Benchmarks › OTIS Mock AIME

OTIS Mock AIME

Older benchmarkBy Epoch AI ↗
Solve mock AIME mathematics problems
100top score · Claude Fable 5
130models ranked
65open-weights models
Accuracyhigher is better
Oct 20, 2026last updated

How the test works

1

Task

45 problems from three OTIS Mock AIME exams

  • What the model doesSolve mock AIME mathematics problems
  • Kinds of tasksCompetition mathematics
2

Run

Step-by-step solve ending ANSWER: X

  • How runs combineMean correctness against the answer key; independent repetitions and sample settings remain run-specific.
3

Score

Accuracy

  • Graded byModel extracts the final integer without seeing the key; deterministic exact-match to the answer key (since September 2025)
  • Better isHigher

OTIS Mock AIME 2024 2025 leaderboardtop 10 of 130

Full leaderboard with filters →
ModelOTIS Mock AIME 2024 2025Per taskPriceSpeedSupportsContext
100high
—$10 / $50—1M
100max
$7.63$10 / $50691M
100max
$5.98$4 / $20961M
100max
$5.46$2 / $101371M
100high
$1.99$2 / $10841M
100max
$1.99$4 / $201141.05M
100max
$3.26$10 / $50691.05M
GPT-6 SolOpenAI
100max
$1.04$2 / $101281.05M
100max
$0.72$2 / $10521.05M
Qwen3.8 Max (0902)Alibaba (Qwen)
100xhigh
—$2 / $6—984K

Versions and boards on Model Menu

BoardModels
130
1

OTIS Mock AIME questions

What is OTIS Mock AIME?›

Solve mock AIME mathematics problems. It has 45 problems from three OTIS Mock AIME exams. Created by Epoch AI.

Which model is best on OTIS Mock AIME?›

Claude Fable 5 leads OTIS Mock AIME 2024 2025 with 100, ahead of Claude Fable 5.1 (100) and Claude Opus 5.5 (100).

What is the best open-weights model on OTIS Mock AIME?›

DeepSeek V4 Pro 0813 with 98.6, then Kimi K3 (97.2).

How is OTIS Mock AIME scored?›

Accuracy; model extracts the final integer without seeing the key; deterministic exact-match to the answer key (since september 2025).

Which versions of OTIS Mock AIME are there?›

OTIS Mock AIME 2024 2025 (130 models), OTIS Mock AIME 2024 2025 (past run) (1 models).