Benchmark Almanac

UniCode

o4-mini 70.3% leads since October 2025
Not in model cards

Models solve generated competitive programming problems using automated test cases. Harbor packages 399 problems with validated reference solutions.

The October 16, 2025 paper compared 19 models on 492 problems. OpenAI o4-mini led that release with a 70.3% pass rate.

Made by
Xinyue Zheng and colleagues
Released
October 2025
Kind
Competitive programming
Level
Competition
Scored
Checked automatically
Answers
Code
Tasks
All public
Measures
Capability
Language
English
In and out
Text → Code
Interaction
Single turn
Items from
Generated
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-5 #71 on the capability index
Last model added
September 2025
Scores refreshed
November 20th, 2026
19
Entries in its paper, 16 models we track

o4-mini leads with 70.3% among models released by September 2025 o4-mini leads with 70.3% among models released by September 2025

Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.

The newest model with a score here came out in September 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 o4-mini high 70.3%
  2. 2 GPT-5 medium 67.7%
  3. 3 Gemini 2.5 Pro Jun 2025 61.6%
  4. 4 DeepSeek-V3.1 thinking 60.5%
  5. 5 deepseek-r1 (snapshot unspecified) 55.6%
  6. 6 o3-mini medium 55.1%
  7. 7 Qwen3-235B-A22B 53.5%
  8. 8 Gemini 2.5 Flash Sep 2025 47.7%
  9. 9 Grok-3 mini 46.4%
  10. 10 Claude 3.7 Sonnet 45.5%
  11. 11 GPT-4.1 mini 42.4%
  12. 12 GPT-4.1 36.5%
  13. 13 Qwen3-Coder-480B-A35B 35.4%
  14. 14 Claude Sonnet 4 32.4%
  15. 15 Llama 4 Maverick 26.2%
  16. 16 gpt-4o (snapshot unspecified) 15.4%
  17. 17 Qwen2.5-Coder-32B 13.4%
  18. 18 Gemma 3 27B 13.1%
  19. 19 llama-3.3-8b-instruct 5.50%

Source: The UniCode paper

#EntryScoreSettingDateSource
1 o4-mini 70.3% high Apr 2025 Official leaderboard
2 GPT-5 67.7% medium Aug 2025 Official leaderboard
3 Gemini 2.5 Pro Jun 2025 61.6% Jun 2025 Official leaderboard
4 DeepSeek-V3.1 60.5% thinking Aug 2025 Official leaderboard
5 deepseek-r1 (snapshot unspecified) 55.6% Official leaderboard
6 o3-mini 55.1% medium Jan 2025 Official leaderboard
7 Qwen3-235B-A22B 53.5% Apr 2025 Official leaderboard
8 Gemini 2.5 Flash Sep 2025 47.7% Sep 2025 Official leaderboard
9 Grok-3 mini 46.4% Apr 2025 Official leaderboard
10 Claude 3.7 Sonnet 45.5% Feb 2025 Official leaderboard
11 GPT-4.1 mini 42.4% Apr 2025 Official leaderboard
12 GPT-4.1 36.5% Apr 2025 Official leaderboard
13 Qwen3-Coder-480B-A35B 35.4% Jul 2025 Official leaderboard
14 Claude Sonnet 4 32.4% May 2025 Official leaderboard
15 Llama 4 Maverick 26.2% Apr 2025 Official leaderboard
16 gpt-4o (snapshot unspecified) 15.4% Official leaderboard
17 Qwen2.5-Coder-32B 13.4% Sep 2024 Official leaderboard
18 Gemma 3 27B 13.1% Mar 2025 Official leaderboard
19 llama-3.3-8b-instruct 5.50% Official leaderboard
#ModelScoreSettingDateSource
1 o4-mini 70.3% high Apr 2025 Official leaderboard
2 GPT-5 67.7% medium Aug 2025 Official leaderboard
3 Gemini 2.5 Pro Jun 2025 61.6% Jun 2025 Official leaderboard
4 DeepSeek-V3.1 60.5% thinking Aug 2025 Official leaderboard
5 o3-mini 55.1% medium Jan 2025 Official leaderboard
6 Qwen3-235B-A22B 53.5% Apr 2025 Official leaderboard
7 Gemini 2.5 Flash Sep 2025 47.7% Sep 2025 Official leaderboard
8 Grok-3 mini 46.4% Apr 2025 Official leaderboard
9 Claude 3.7 Sonnet 45.5% Feb 2025 Official leaderboard
10 GPT-4.1 mini 42.4% Apr 2025 Official leaderboard
11 GPT-4.1 36.5% Apr 2025 Official leaderboard
12 Qwen3-Coder-480B-A35B 35.4% Jul 2025 Official leaderboard
13 Claude Sonnet 4 32.4% May 2025 Official leaderboard
14 Llama 4 Maverick 26.2% Apr 2025 Official leaderboard
15 Qwen2.5-Coder-32B 13.4% Sep 2024 Official leaderboard
16 Gemma 3 27B 13.1% Mar 2025 Official leaderboard

Rankings by part

We read 3 parts from the benchmark’s paper: Easy, Medium and Hard. Each ranks models on one part of the test. o4-mini leads every one of them. Arrows show how far a model moved from its overall rank among the models we track.

Easy

16 models we track have a score here.

  1. 1 o4-mini 94.9%
  2. 2 Gemini 2.5 Pro Jun 2025 ↑1 94%
  3. 3 GPT-5 ↓1 89.5%
  4. 4 DeepSeek-V3.1 89.2%
  5. 5 o3-mini 86.2%
  6. 6 Gemini 2.5 Flash Sep 2025 ↑1 81.4%
  7. 7 Qwen3-235B-A22B ↓1 80.2%
  8. 8 Grok-3 mini 77.8%
  9. 9 Claude 3.7 Sonnet 76.2%
  10. 10 GPT-4.1 mini 73.7%

The best score, 70.3%, costs $0.03 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Meta xAI OpenAI DeepSeek Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.0001 $0.001 $0.01 $0.10 $1.00 Cost per task, log scale o4-mini, high o4-mini, medium DeepSeek-V3.1 Llama 4 Maverick 0% 20% 40% 60% 80% 100% $0.0001 $0.001 $0.01 $0.10 $1.00 Cost per task, log scale o4-mini, high DeepSeek-V3.1

Source: The UniCode paper

ModelEffortHarnessScoreCost per taskOutput tokens
o4-mini high 70.3% $0.03
GPT-5 medium 67.7% $0.04
o4-mini medium 66.1% $0.02
Gemini 2.5 Pro Jun 2025 61.6% $0.20
DeepSeek-V3.1 thinking 60.5% $0.03
o3-mini medium 55.1% $0.02
Qwen3-235B-A22B 53.5% $0.03
DeepSeek-V3.1 49.8% $0.007
Gemini 2.5 Flash Sep 2025 47.7% $0.009
Grok-3 mini 46.4% $0.004
Claude 3.7 Sonnet 45.5% $0.13
GPT-4.1 mini 42.4% $0.007
GPT-4.1 36.5% $0.007
Qwen3-Coder-480B-A35B 35.4% $0.01
Claude Sonnet 4 32.4% $0.02
Llama 4 Maverick 26.2% $0.0006
Qwen2.5-Coder-32B 13.4% $0.004
Gemma 3 27B 13.1%

The best score, 70.3%, costs $0.03 a task

Runs on the cost frontier score higher than any run that's cheaper. The costs are the leaderboard's own estimates, so compare them only within this benchmark.

Meta xAI OpenAI DeepSeek Cost frontier One model at each effort level
0% 20% 40% 60% 80% 100% $0.0001 $0.001 $0.01 $0.10 $1.00 Cost per task, log scale o4-mini, high o4-mini, medium DeepSeek-V3.1 Llama 4 Maverick 0% 20% 40% 60% 80% 100% $0.0001 $0.001 $0.01 $0.10 $1.00 Cost per task, log scale o4-mini, high DeepSeek-V3.1

Source: The UniCode paper

ModelEffortHarnessScoreCost per taskOutput tokens
o4-mini high 70.3% $0.03
GPT-5 medium 67.7% $0.04
o4-mini medium 66.1% $0.02
Gemini 2.5 Pro Jun 2025 61.6% $0.20
DeepSeek-V3.1 thinking 60.5% $0.03
o3-mini medium 55.1% $0.02
Qwen3-235B-A22B 53.5% $0.03
DeepSeek-V3.1 49.8% $0.007
Gemini 2.5 Flash Sep 2025 47.7% $0.009
Grok-3 mini 46.4% $0.004
Claude 3.7 Sonnet 45.5% $0.13
GPT-4.1 mini 42.4% $0.007
GPT-4.1 36.5% $0.007
Qwen3-Coder-480B-A35B 35.4% $0.01
Claude Sonnet 4 32.4% $0.02
Llama 4 Maverick 26.2% $0.0006
Qwen2.5-Coder-32B 13.4% $0.004
Gemma 3 27B 13.1%

Number one over time

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

OpenAI
Benchmark paper Not tested since Oct 2025
o4-mini 70.3%

Source: The UniCode paper

Scores over time

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

Plot with other benchmarks
In the benchmark’s paper Best so far
0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) o4-mini o3-mini Qwen2.5-Coder-32B 0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) o4-mini o3-mini

Source: Official UniCode leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Coding

All 523 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%
  • 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%
  • SWE-bench Multilingual In use 300 GitHub-issue tasks across nine programming languages in SWE-bench format. Patch graded by repository tests. Claude Fable 5.2 96.6%
  • 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

Data audit

9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 14 months ago, in September 2025, for Gemini 2.5 Flash. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the benchmark’s paper (19 entries).
  • Pass Models scored 16 models we track, plus 3 other entries on its leaderboard.
  • Warning Newest score The newest came 14 months ago, in September 2025, for Gemini 2.5 Flash.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5 (#71).
  • Pass Headroom The best is 70.3%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 18 rows for models we track and 3 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check