Benchmark Almanac

IOI v2 (Vals)

GPT-6 Astra 100% leads since Oct 2026 Not in model cards

Agents solve 18 International Olympiad in Informatics problems from 2024, 2025 and 2026 in C++ inside OpenCode. Official tests and graders score the final solution by subtask after the run.

Unlike v1 there is no interactive grading tool, and Vals says the two harnesses are not comparable. Score is the mean of three yearly scores. Released between 2026-08-09 and 2026-10-29.

Made by Vals AI Released Oct 2026 Competitive programming All tasks public Last model added Oct 2026 Scores refreshed 2 Nov 2026
Level
Competition The strongest students, in olympiad-style contests
Scored
Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
Answers
Code A program or patch run against tests.
Leaderboard vals.ai
45
Models scored on its own leaderboard

2 models share the top score of 100%

Best score per model from the benchmark's own leaderboard, read on 2 Nov 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.

45 models. Scroll sideways for the rest.

  1. 100.0 GPT-6 Astra
  2. 100.0 Gemini 4 Argon
  3. 96.9 GPT-6.1 Sol
  4. 95.1 Claude Opus 5.5
  5. 91.2 GPT-5.6 Sol
  6. 90.8 Claude Fable 5.1
  7. 87.6 GPT-5.6 Terra
  8. 84.3 Claude Opus 5
  9. 83.1 Claude Sonnet 5.5
  10. 82.6 GPT-6 Sol
  11. 81.9 Kimi K3.1 Pro
  12. 79.8 Muse Wonder Full
  13. 68.9 Qwen 3.8 Max
  14. 68.4 GLM-5.3
  15. 67.8 Gemini 3.7 Flash
  16. 66.9 Muse Wonder
  17. 61.8 GPT-5.6 Luna
  18. 59.3 Tencent Hy4 preview
  19. 57.7 Grok 4.7
  20. 56.9 Gemini 3.8 Flash
  21. 56.6 Muse Spark 1.3
  22. 55.6 GPT-6 Luna
  23. 53.8 GPT-5.3 Codex
  24. 52.5 GLM-5.3-Flash
  25. 51.8 Gemini 3.1 Pro
  26. 51.6 DeepSeek V4 Pro 0813
  27. 48.9 Kimi K3
  28. 47.7 MiMo-V2.6 Flash
  29. 47.6 Grok 4.6
  30. 47.3 Claude Haiku 5.5
  31. 46.3 Ember-1
  32. 45.3 Mistral Large 4
  33. 45.0 Claude Sonnet 5
  34. 40.6 Grok 4.5
  35. 40.3 DeepSeek V4.1 Flash
  36. 39.3 MiMo-V2.6 Pro
  37. 39.1 Qwen 3.8 27B
  38. 38.1 Step 5 Preview
  39. 35.1 Gemini 3.6 Flash
  40. 32.7 DeepSeek V4 Flash 0731
  41. 22.8 Beam 1.0
  42. 21.8 Muse Spark 1.2
  43. 14.9 Inkling
  44. 9.33 Inkling-Small
  45. 2.44 Mercury 2.5
  1. 1 GPT-6 Astra 100%
  2. 2 Gemini 4 Argon 100%
  3. 3 GPT-6.1 Sol 96.9%
  4. 4 Claude Opus 5.5 95.1%
  5. 5 GPT-5.6 Sol 91.2%
  6. 6 Claude Fable 5.1 90.8%
  7. 7 GPT-5.6 Terra 87.6%
  8. 8 Claude Opus 5 84.3%
  9. 9 Claude Sonnet 5.5 83.1%
  10. 10 GPT-6 Sol 82.6%
  11. 11 Kimi K3.1 Pro 81.9%
  12. 12 Muse Wonder Full 79.8%
  13. 13 Qwen 3.8 Max 68.9%
  14. 14 GLM-5.3 68.4%
  15. 15 Gemini 3.7 Flash 67.8%
  16. 16 Muse Wonder 66.9%
  17. 17 GPT-5.6 Luna 61.8%
  18. 18 Tencent Hy4 preview 59.3%
  19. 19 Grok 4.7 57.7%
  20. 20 Gemini 3.8 Flash 56.9%
  21. 21 Muse Spark 1.3 max 56.6%
  22. 22 GPT-6 Luna 55.6%
  23. 23 GPT-5.3 Codex 53.8%
  24. 24 GLM-5.3-Flash 52.5%
  25. 25 Gemini 3.1 Pro 51.8%
  26. 26 DeepSeek V4 Pro 0813 51.6%
  27. 27 Kimi K3 48.9%
  28. 28 MiMo-V2.6 Flash 47.7%
  29. 29 Grok 4.6 47.6%
  30. 30 Claude Haiku 5.5 47.3%
  31. 31 Ember-1 46.3%
  32. 32 Mistral Large 4 45.3%
  33. 33 Claude Sonnet 5 45%
  34. 34 Grok 4.5 40.6%
  35. 35 DeepSeek V4.1 Flash 40.3%
  36. 36 MiMo-V2.6 Pro 39.3%
  37. 37 Qwen 3.8 27B 39.1%
  38. 38 Step 5 Preview 38.1%
  39. 39 Gemini 3.6 Flash 35.1%
  40. 40 DeepSeek V4 Flash 0731 32.7%
  41. 41 Beam 1.0 22.8%
  42. 42 Muse Spark 1.2 21.8%
  43. 43 Inkling 14.9%
  44. 44 Inkling-Small 9.33%
  45. 45 Mercury 2.5 2.44%
#ModelScoreSettingDateSource
1 GPT-6 Astra 100% Sep 2026 Official leaderboard
2 Gemini 4 Argon 100% Sep 2026 Official leaderboard
3 GPT-6.1 Sol 96.9% Sep 2026 Official leaderboard
4 Claude Opus 5.5 95.1% Sep 2026 Official leaderboard
5 GPT-5.6 Sol 91.2% Jul 2026 Official leaderboard
6 Claude Fable 5.1 90.8% Sep 2026 Official leaderboard
7 GPT-5.6 Terra 87.6% Jul 2026 Official leaderboard
8 Claude Opus 5 84.3% Jul 2026 Official leaderboard
9 Claude Sonnet 5.5 83.1% Sep 2026 Official leaderboard
10 GPT-6 Sol 82.6% Sep 2026 Official leaderboard
11 Kimi K3.1 Pro 81.9% Oct 2026 Official leaderboard
12 Muse Wonder Full 79.8% Oct 2026 Official leaderboard
13 Qwen 3.8 Max 68.9% Aug 2026 Official leaderboard
14 GLM-5.3 68.4% Aug 2026 Official leaderboard
15 Gemini 3.7 Flash 67.8% Aug 2026 Official leaderboard
16 Muse Wonder 66.9% Oct 2026 Official leaderboard
17 GPT-5.6 Luna 61.8% Jul 2026 Official leaderboard
18 Tencent Hy4 preview 59.3% Aug 2026 Official leaderboard
19 Grok 4.7 57.7% Sep 2026 Official leaderboard
20 Gemini 3.8 Flash 56.9% Sep 2026 Official leaderboard
21 Muse Spark 1.3 56.6% max Sep 2026 Official leaderboard
22 GPT-6 Luna 55.6% Sep 2026 Official leaderboard
23 GPT-5.3 Codex 53.8% Feb 2026 Official leaderboard
24 GLM-5.3-Flash 52.5% Aug 2026 Official leaderboard
25 Gemini 3.1 Pro 51.8% Feb 2026 Official leaderboard
26 DeepSeek V4 Pro 0813 51.6% Aug 2026 Official leaderboard
27 Kimi K3 48.9% Jul 2026 Official leaderboard
28 MiMo-V2.6 Flash 47.7% Sep 2026 Official leaderboard
29 Grok 4.6 47.6% Aug 2026 Official leaderboard
30 Claude Haiku 5.5 47.3% Oct 2026 Official leaderboard
31 Ember-1 46.3% Sep 2026 Official leaderboard
32 Mistral Large 4 45.3% Oct 2026 Official leaderboard
33 Claude Sonnet 5 45% Jun 2026 Official leaderboard
34 Grok 4.5 40.6% Jul 2026 Official leaderboard
35 DeepSeek V4.1 Flash 40.3% Sep 2026 Official leaderboard
36 MiMo-V2.6 Pro 39.3% Sep 2026 Official leaderboard
37 Qwen 3.8 27B 39.1% Aug 2026 Official leaderboard
38 Step 5 Preview 38.1% Sep 2026 Official leaderboard
39 Gemini 3.6 Flash 35.1% Jul 2026 Official leaderboard
40 DeepSeek V4 Flash 0731 32.7% Jul 2026 Official leaderboard
41 Beam 1.0 22.8% Oct 2026 Official leaderboard
42 Muse Spark 1.2 21.8% Aug 2026 Official leaderboard
43 Inkling 14.9% Jul 2026 Official leaderboard
44 Inkling-Small 9.33% Jul 2026 Official leaderboard
45 Mercury 2.5 2.44% Sep 2026 Official leaderboard

The best score, 100%, costs $5.31 a task

Runs on the cost frontier score higher than any run that's cheaper. GPT-6.1 Sol comes within 3.1 points of the best for $1.26 a task, 4 times cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

OpenAI Inception Labs Google DeepMind Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.10 $1.00 $10 $100 Cost per task, log scale Gemini 4 Argon GPT-6.1 Sol GPT-5.6 Luna GPT-6 Luna Mercury 2.5 0% 20% 40% 60% 80% 100% $0.10 $1.00 $10 $100 Cost per task, log scale Gemini 4 Argon GPT-5.6 Luna
ModelEffortHarnessScoreCost per taskOutput tokens
GPT-6 Astra 100% $6.50
Gemini 4 Argon 100% $5.31
GPT-6.1 Sol 96.9% $1.26
Claude Opus 5.5 95.1% $5.26
GPT-5.6 Sol 91.2% $7.98
Claude Fable 5.1 90.8% $11
GPT-5.6 Terra 87.6% $8.66
Claude Opus 5 84.3% $16
Claude Sonnet 5.5 83.1% $7.52
GPT-6 Sol 82.6% $2.70
Kimi K3.1 Pro 81.9% $5.02
Muse Wonder Full 79.8% $7.08
Qwen 3.8 Max 68.9% $9.47
GLM-5.3 68.4% $7.67
Gemini 3.7 Flash 67.8% $3.40
Muse Wonder 66.9% $4.54
GPT-5.6 Luna 61.8% $0.58
Tencent Hy4 preview 59.3% $1.35
Grok 4.7 57.7% $13
Gemini 3.8 Flash 56.9% $3.98
Muse Spark 1.3 max 56.6% $1.73
GPT-6 Luna 55.6% $0.20
GPT-5.3 Codex 53.8% $2.68
GLM-5.3-Flash 52.5% $0.43
Gemini 3.1 Pro 51.8% $1.25
DeepSeek V4 Pro 0813 51.6% $2.15
Kimi K3 48.9% $14
MiMo-V2.6 Flash 47.7% $0.22
Grok 4.6 47.6% $8.18
Claude Haiku 5.5 47.3% $0.83
Ember-1 46.3% $24
Mistral Large 4 45.3% $2.73
Claude Sonnet 5 45% $13
Muse Spark 1.3 43.9% $2.36
Grok 4.5 40.6% $5.64
DeepSeek V4.1 Flash 40.3% $0.24
MiMo-V2.6 Pro 39.3% $0.84
Qwen 3.8 27B 39.1% $3.55
Step 5 Preview 38.1% $2.03
Gemini 3.6 Flash 35.1% $8.67
DeepSeek V4 Flash 0731 32.7% $0.37
Beam 1.0 22.8%
Muse Spark 1.2 21.8% $2.64
Inkling 14.9% $1.71
Inkling-Small 9.33% $0.34
Mercury 2.5 2.44% $0.16

Number one over time

Lab claims count from the day the card came out, independent runs from the model’s release.

OpenAI
Own leaderboard Passed 85% in Oct 2026, by GPT-6 Astra
GPT-6 Astra 100%

The best score passed 85% in Jul 2026

All the scores we have, one dot each, so a model tested at several settings appears more than once.

On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) GPT-6 Astra GPT-5.6 Sol GPT-5.3 Codex 0% 20% 40% 60% 80% 100% 2026 2027 Near the ceiling (85%) GPT-6 Astra GPT-5.6 Sol

No lab model card reports it yet

We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark.

Versions

We switch the top row to each new version when it comes out. Open it for one row per version. An old version stays in play while people still test it, so whoever first takes it past 85% gets credit for closing it out.

xAI Google DeepMind OpenAI Anthropic
IOI v2 (Vals) Passed 85% in Oct 2026, by GPT-6 Astra
GPT-6 Astra 100%
IOI (Vals) Passed 85% in Jul 2026, by GPT-5.6 Sol
Claude Opus 5 91.7%
IOI v2 (Vals) Passed 85% in Oct 2026, by GPT-6 Astra
GPT-6 Astra 100%
Replaces IOI (Vals)

More in Coding

All 243 in Coding
  • SWE-bench Verified In use 500 human-validated GitHub issues from 12 Python repos. Agent writes a patch that must pass hidden tests. Claude Opus 4.7 83.5%
  • SWE-bench Multilingual In use 300 GitHub-issue tasks across nine programming languages in SWE-bench format. Patch graded by repository tests. Claude Opus 5.5 93.9%
  • LiveCodeBench v6 In use Competitive programming problems collected after model cutoffs from LeetCode, AtCoder, Codeforces. Scored pass@1 on hidden tests. DeepSeek-V4-Pro 93.5%
  • DeepSWE In use 113 original long-horizon engineering tasks across 91 repos and five languages, written from scratch. Hand-written verifiers accept alternative solutions. GPT-6 Astra 74.1%
  • SciCode In use Scientist-curated research coding problems split into subproblems across natural sciences. Code checked against test cases. Seed2.1 Pro 59.8%
  • Codeforces rating In use Model solves recent Codeforces contest problems. An Elo-style rating is computed from simulated contest performance. DeepSeek V4.1 Flash 3,471