BenchLeader
Menu
Best LLMsChoose a modelLeaderboardsModelsBenchmarksProvidersReleases12

IOI

International Olympiad in Informatics problems. Run by Vals AI.

As of 24 Oct 2026, Gemini 4 Argon leads IOI on BenchLeader with 100.0%, ahead of GPT-6 Astra at 100.0%, across 43 model configurations with a published result.

Published by
Vals AI
Category
Coding
Index weight
0.5
Models
43
Data as of
24 Oct 2026
Newest result
20 Oct 2026

Vals AI (vals.ai).

What the test looks like

Problems from the International Olympiad in Informatics, the hardest school-level programming competition, solved under the official test cases.

How it is scored

Percent of the available points scored across problems, run by Vals AI.

What to keep in mind

A handful of problems per year, each partially scored, so a single problem's subtasks move the total. It measures algorithmic problem solving, not software engineering.

Top 15
  1. 1Gemini 4 Argon (high)100.0%
  2. 2GPT-6 Astra (max)100.0%
  3. 3GPT-6.1 Sol (max)96.9%
  4. 4Claude Opus 5.595.1%
  5. 5GPT-5.6 Sol (max)91.2%
  6. 6Claude Fable 5.190.8%
  7. 7GPT-5.6 Terra (max)87.6%
  8. 8Claude Opus 584.3%
  9. 9Claude Sonnet 5.583.1%
  10. 10GPT-6 Sol (max)82.6%
  11. 11Muse Wonder Full (max)79.8%
  12. 12Qwen3.8 Max (max)68.9%
  13. 13GLM 5.3 (max)68.4%
  14. 14Gemini 3.7 Flash (high)67.8%
  15. 15Muse Wonder (max)66.9%
43 of 43
#
1Gemini 4 Argonhigh100.0%70.82026-10-20
2GPT-6 Astramax100.0%70.42026-10-20
3GPT-6.1 Solmax96.9%69.92026-10-20
4Claude Opus 5.595.1%62.02026-10-20
5GPT-5.6 Solmax91.2%68.02026-10-20
6Claude Fable 5.190.8%66.42026-10-20
7GPT-5.6 Terramax87.6%64.42026-10-20
8Claude Opus 584.3%65.52026-10-20
9Claude Sonnet 5.583.1%–2026-10-20
10GPT-6 Solmax82.6%66.42026-10-20
11Muse Wonder Fullmax79.8%70.72026-10-20
12Qwen3.8 Maxmax68.9%63.32026-10-20
13GLM 5.3max68.4%63.52026-10-20
14Gemini 3.7 Flashhigh67.8%63.52026-10-20
15Muse Wondermax66.9%67.52026-10-20
16GPT-5.6 Lunamax61.8%59.92026-10-20
17Grok 4.7xhigh57.7%61.62026-10-20
18Gemini 3.8 Flashhigh56.9%64.42026-10-20
19Muse Spark 1.3max56.6%67.12026-10-20
20GPT-6 Lunamax55.6%60.02026-10-20
21GPT-5.3-Codexxhigh53.8%63.82026-10-20
22GLM 5.3 Flashmax52.5%54.22026-10-20
23Gemini 3.1 Prohigh51.8%60.32026-10-20
24DeepSeek V4 Promax51.6%60.82026-10-20
25Kimi K3max48.9%64.72026-10-20
26MiMo-V2.6-Flash47.7%59.72026-10-20
27Grok 4.6high47.6%62.92026-10-20
28Claude Haiku 5.547.3%–2026-10-20
29Ember-146.3%–2026-10-20
30Mistral Large 4high45.3%49.82026-10-20
31Claude Sonnet 545.0%56.12026-10-20
32Muse Spark 1.3xhigh43.9%64.92026-10-20
33Grok 4.5high40.6%59.32026-10-20
34DeepSeek V4.1 Flashmax40.3%61.22026-10-20
35MiMo-V2.6-Pro39.3%64.02026-10-20
36Qwen3.8 27Bxhigh39.1%56.22026-10-20
37Step 5 Preview38.1%60.42026-10-20
38Gemini 3.6 Flashhigh35.1%62.02026-10-20
39DeepSeek V4 Flashhigh32.7%57.42026-10-20
40NewBeam 1.022.8%51.12026-10-20
41Muse Spark 1.2xhigh21.8%63.22026-10-20
42Inkling Small9.3%54.82026-10-20
43Mercury 2.5high2.4%–2026-10-20

More coding benchmarks · All LLM benchmarks

Cite as: BenchLeader, “IOI leaderboard”, data as of 24 Oct 2026, https://www.benchleader.com/benchmarks/vals_ioi.

IOI: questions

What does IOI measure?
Problems from the International Olympiad in Informatics, the hardest school-level programming competition, solved under the official test cases. Scores are reported in percent of tasks solved; higher is better.
Which AI model leads IOI?
Gemini 4 Argon leads IOI with 100.0% as of 24 Oct 2026, ahead of GPT-6 Astra at 100.0%.
How many models have IOI results?
43 model configurations have a IOI result on BenchLeader, all taken from Vals AI.
Who runs IOI and how often is it updated?
IOI is published by Vals AI. BenchLeader re-reads the published results once a day and records the date each result was published.
Does IOI count toward the BenchLeader Index?
Yes. IOI contributes to the coding category of the BenchLeader Index, normalised so that 50 is the average of the evaluated models.