BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases9

OTIS Mock AIME

Mock AIME competition problems from 2024–2025. Run by Epoch AI.

As of 26 Oct 2026, Claude Opus 5.5 leads OTIS Mock AIME on BenchLeader with 100.0%, ahead of Gemini 4 Argon at 100.0%, across 276 model configurations with a published result.

Published by
Epoch AI Benchmarking Hub
Category
Maths
Index weight
0.5
Models
276
Data as of
26 Oct 2026
Newest result
17 Oct 2026

CC BY 4.0 — Epoch AI, ‘AI Benchmarking Hub’, epoch.ai/benchmarks. Mirrored boards credit their original publishers.

What the test looks like

Mock AIME papers written by Evan Chen's OTIS programme for the 2024 and 2025 seasons: fifteen-problem competition sets in the style of the American Invitational Mathematics Examination, where every answer is an integer from 0 to 999.

How it is scored

Percent of problems answered correctly across the papers, run by Epoch AI with a fixed prompt and several attempts per problem.

What to keep in mind

The problems were written for humans but are public, so later models may have seen them. Competition maths rewards a particular kind of trick-finding; strong scores here do not necessarily transfer to proofs or applied problems.

Top 15
  1. 1Claude Opus 5.5 (max)100.0%
  2. 2Gemini 4 Argon (high)100.0%
  3. 3Claude Fable 5.1 (max)100.0%
  4. 4GPT-6 Astra (max)100.0%
  5. 5GPT-6.1 Sol (max)100.0%
  6. 6Claude Sonnet 5.5 (max)100.0%
  7. 7GPT-5.6 Sol (max)100.0%
  8. 8GPT-6 Sol (max)100.0%
  9. 9Claude Fable 5 (high)100.0%
  10. 10Qwen3.8 Max (xhigh)100.0%
  11. 11GPT-5.5 Pre Release (xhigh)100.0%
  12. 12GPT-5.5 Pro Pre Release (xhigh)100.0%
  13. 13Claude Fable 5 (max)99.7%
  14. 14GPT-5.6 Terra (max)99.7%
  15. 15Muse Wonder (max)99.4%
276 of 276
#
1Claude Opus 5.5max100.0%72.62026-09-22
2Gemini 4 Argonhigh100.0%70.72026-09-30
3Claude Fable 5.1max100.0%70.52026-09-01
4GPT-6 Astramax100.0%70.42026-09-03
5GPT-6.1 Solmax100.0%69.92026-09-29
6Claude Sonnet 5.5max100.0%69.42026-09-28
7GPT-5.6 Solmax100.0%68.02026-07-09
8GPT-6 Solmax100.0%66.42026-09-22
9Claude Fable 5high100.0%64.42026-06-09
10Qwen3.8 Maxxhigh100.0%57.12026-09-01
11GPT-5.5 Pre Releasexhigh100.0%–2026-04-23
12GPT-5.5 Pro Pre Releasexhigh100.0%–2026-04-23
13Claude Fable 5max99.7%69.32026-06-09
14GPT-5.6 Terramax99.7%64.42026-07-09
15Muse Wondermax99.4%67.52026-10-12
16DeepSeek V4.1 Flash 1017max99.4%64.02026-10-17
17Muse Spark 1.3xhigh99.2%64.92026-09-02
18Grok 4.6xhigh99.2%62.52026-08-12
19Claude Opus 5max98.9%68.82026-07-24
20Gemini 3.8 Flashhigh98.9%64.42026-09-02
21Claude Haiku 5.5xhigh98.9%61.52026-10-07
22GPT-6 Lunamax98.9%60.02026-09-22
23DeepSeek V4 Promax98.6%60.82026-08-13
24Claude Opus 4.8max98.3%63.32026-05-28
25DeepSeek V4.1 Flashmax98.3%61.22026-09-09
26GPT-5.6 Lunamax98.3%59.92026-07-09
27Mistral Large 4high98.3%49.82026-10-06
28Grok 4.7xhigh98.1%61.62026-09-21
29Muse Spark Open 816B A42Bxhigh98.1%58.62026-10-15
30Claude Opus 4.7xhigh97.8%57.42026-04-16
31Claude Opus 597.8%65.12026-07-24
32Grok 4.6high97.8%62.92026-08-12
33Grok 4.5high97.8%59.32026-07-08
34GPT-5.4high97.8%58.92026-03-05
35Claude Fable 5low97.8%–2026-06-09
36Claude Opus 4.8low97.8%–2026-05-28
37Kimi K3max97.2%64.82026-07-16
38Gemini 3.7 Flashhigh97.2%63.52026-08-13
39Claude Haiku 5.5high97.2%59.62026-10-07
40GPT-5.2high96.1%56.02025-12-11
41Kimi K2.696.1%55.32026-04-20
42GPT-5.2xhigh96.1%60.72025-12-11
43Gemini 3.1 Pro95.6%61.82026-02-19
44Gemini 3.5 Flashhigh95.6%63.12026-05-19
45GPT-5.6 Sollow95.6%62.22026-07-09
46Qwen3.7max95.6%60.62026-05-19
47Gemini 3.1 Prohigh95.6%60.32026-02-19
48Gemini 3 Flashhigh95.6%58.32025-12-17
49DeepSeek V4 Prohigh95.6%57.52026-04-24
50Kimi K2.7 Code95.6%55.82026-06-12
51GPT-5.4medium95.6%55.02026-03-05
52GPT-5.4xhigh95.3%64.72026-03-05
53Claude Sonnet 5xhigh94.7%59.02026-06-30
54Claude Opus 4.694.4%62.52026-02-05
55DeepSeek V4 Flashmax94.4%59.42026-07-31
56Gemini 3.6 Flashhigh94.2%62.02026-07-21
57GPT-5.2medium93.9%57.62025-12-11
58GLM 5.3 Flashmax93.9%54.22026-08-20
59Claude Opus 5low93.3%61.32026-07-24
60Qwen3.6 Plus93.3%58.82026-03-31

More maths benchmarks · All LLM benchmarks

Cite as: BenchLeader, “OTIS Mock AIME leaderboard”, data as of 26 Oct 2026, https://www.benchleader.com/benchmarks/otis_mock_aime.

OTIS Mock AIME: questions

What does OTIS Mock AIME measure?
Mock AIME papers written by Evan Chen's OTIS programme for the 2024 and 2025 seasons: fifteen-problem competition sets in the style of the American Invitational Mathematics Examination, where every answer is an integer from 0 to 999. Scores are reported in percent of tasks solved; higher is better.
Which AI model leads OTIS Mock AIME?
Claude Opus 5.5 leads OTIS Mock AIME with 100.0% as of 26 Oct 2026, ahead of Gemini 4 Argon at 100.0%.
How many models have OTIS Mock AIME results?
276 model configurations have a OTIS Mock AIME result on BenchLeader, all taken from Epoch AI Benchmarking Hub.
Who runs OTIS Mock AIME and how often is it updated?
OTIS Mock AIME is published by Epoch AI Benchmarking Hub. BenchLeader re-reads the published results once a day and records the date each result was published.
Does OTIS Mock AIME count toward the BenchLeader Index?
Yes. OTIS Mock AIME contributes to the maths category of the BenchLeader Index, normalised so that 50 is the average of the evaluated models.