Benchmark Almanac

CRUST-bench

o3 48% leads since April 2025
Not in model cards

Agents translate 100 C projects into safe Rust. Tests check whether each translation matches the supplied Rust interface and preserves the original behavior.

The Harbor comparison available November 19, 2026 reports one model. GPT-5 nano with Codex solved 58.3% on both the original benchmark and Harbor.

Released
April 2025
Kind
Fixing real repositories
Level
Expert
Scored
Checked automatically
Answers
Agent task
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Code → Code
Interaction
Agentic
Environment
Terminal
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
In views
Agents
Strongest model tested
o3 #84 on the capability index
Last model added
May 2025
Scores refreshed
November 20th, 2026
12
Entries in its paper, 7 models we track

o3 leads with 48% among models released by May 2025 o3 leads with 48% among models released by May 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 May 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 o3 Test repair (3 rounds) 48%
  2. 2 Claude Opus 4 Test repair (3 rounds) 40%
  3. 3 o1 Test repair (3 rounds) 37%
  4. 4 Claude 3.7 Sonnet Test repair (3 rounds) 32%
  5. 5 Claude 3.5 (snapshot unspecified) Test repair (3 rounds) 24%
  6. 6 GPT-4o (snapshot unspecified) Test repair (3 rounds) 22%
  7. 7 o1-mini Test repair (3 rounds) 21%
  8. 8 Gemini 1.5 Pro (snapshot unspecified) Test repair (3 rounds) 14%
  9. 9 Virtuoso (Distilled Deepseek V3) Compiler repair (3 rounds) 6.00%
  10. 10 Deepseek-Coder-32B Single-shot 0%
  11. 11 Qwen2.5-Coder-32B-Instruct Single-shot 0%
  12. 12 QwQ-32B-Preview Single-shot 0%

Source: The CRUST-bench paper

#EntryScoreSettingDateSource
1 o3 48% Test repair (3 rounds) Apr 2025 Official leaderboard
2 Claude Opus 4 40% Test repair (3 rounds) May 2025 Official leaderboard
3 o1 37% Test repair (3 rounds) Dec 2024 Official leaderboard
4 Claude 3.7 Sonnet 32% Test repair (3 rounds) Feb 2025 Official leaderboard
5 Claude 3.5 (snapshot unspecified) 24% Test repair (3 rounds) Official leaderboard
6 GPT-4o (snapshot unspecified) 22% Test repair (3 rounds) Official leaderboard
7 o1-mini 21% Test repair (3 rounds) Sep 2024 Official leaderboard
8 Gemini 1.5 Pro (snapshot unspecified) 14% Test repair (3 rounds) Official leaderboard
9 Virtuoso (Distilled Deepseek V3) 6.00% Compiler repair (3 rounds) Official leaderboard
10 Deepseek-Coder-32B 0% Single-shot Official leaderboard
11 Qwen2.5-Coder-32B-Instruct 0% Single-shot Nov 2024 Official leaderboard
12 QwQ-32B-Preview 0% Single-shot Nov 2024 Official leaderboard
#ModelScoreSettingDateSource
1 o3 48% Test repair (3 rounds) Apr 2025 Official leaderboard
2 Claude Opus 4 40% Test repair (3 rounds) May 2025 Official leaderboard
3 o1 37% Test repair (3 rounds) Dec 2024 Official leaderboard
4 Claude 3.7 Sonnet 32% Test repair (3 rounds) Feb 2025 Official leaderboard
5 o1-mini 21% Test repair (3 rounds) Sep 2024 Official leaderboard
6 Qwen2.5-Coder-32B-Instruct 0% Single-shot Nov 2024 Official leaderboard
7 QwQ-32B-Preview 0% Single-shot Nov 2024 Official leaderboard

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
o3
Single-shot Compiler re… Test repair…
48%
Claude Opus 4
Single-shot Compiler re… Test repair…
40%
o1
Single-shot Compiler re… Test repair…
37%
Claude 3.7 Sonnet
Single-shot Compiler re… Test repair…
32%
o1-mini
Single-shot Compiler re… Test repair…
21%
Qwen2.5-Coder-32B-Instruct
Single-shot
0%
QwQ-32B-Preview
Single-shot
0%

Source: The CRUST-bench paper

ModelEffortHarnessScoreCost per taskOutput tokens
o3 Test repair (3 rounds) 48%
Claude Opus 4 Test repair (3 rounds) 40%
o1 Test repair (3 rounds) 37%
Claude 3.7 Sonnet Test repair (3 rounds) 32%
Claude 3.7 Sonnet Adapted SWE-agent 32%
o3 Compiler repair (3 rounds) 31%
Claude Opus 4 Compiler repair (3 rounds) 29%
o1 Compiler repair (3 rounds) 28%
Claude 3.7 Sonnet Compiler repair (3 rounds) 23%
Claude Opus 4 Single-shot 22%
o1-mini Test repair (3 rounds) 21%
o3 Single-shot 19%
o1-mini Compiler repair (3 rounds) 16%
o1 Single-shot 15%
Claude 3.7 Sonnet Single-shot 13%
o1-mini Single-shot 9.00%
Qwen2.5-Coder-32B-Instruct Single-shot 0%
Qwen2.5-Coder-32B-Instruct Compiler repair (3 rounds) 0%
Qwen2.5-Coder-32B-Instruct Test repair (3 rounds) 0%
QwQ-32B-Preview Single-shot 0%
QwQ-32B-Preview Compiler repair (3 rounds) 0%
QwQ-32B-Preview Test repair (3 rounds) 0%

Scores at each effort level

Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.

Lowest effort to highest, light to dark
o3
Single-shot Compiler re… Test repair…
48%
Claude Opus 4
Single-shot Compiler re… Test repair…
40%
o1
Single-shot Compiler re… Test repair…
37%
Claude 3.7 Sonnet
Single-shot Compiler re… Test repair…
32%
o1-mini
Single-shot Compiler re… Test repair…
21%
Qwen2.5-Coder-32B-Instruct
Single-shot
0%
QwQ-32B-Preview
Single-shot
0%

Source: The CRUST-bench paper

ModelEffortHarnessScoreCost per taskOutput tokens
o3 Test repair (3 rounds) 48%
Claude Opus 4 Test repair (3 rounds) 40%
o1 Test repair (3 rounds) 37%
Claude 3.7 Sonnet Test repair (3 rounds) 32%
Claude 3.7 Sonnet Adapted SWE-agent 32%
o3 Compiler repair (3 rounds) 31%
Claude Opus 4 Compiler repair (3 rounds) 29%
o1 Compiler repair (3 rounds) 28%
Claude 3.7 Sonnet Compiler repair (3 rounds) 23%
Claude Opus 4 Single-shot 22%
o1-mini Test repair (3 rounds) 21%
o3 Single-shot 19%
o1-mini Compiler repair (3 rounds) 16%
o1 Single-shot 15%
Claude 3.7 Sonnet Single-shot 13%
o1-mini Single-shot 9.00%
Qwen2.5-Coder-32B-Instruct Single-shot 0%
Qwen2.5-Coder-32B-Instruct Compiler repair (3 rounds) 0%
Qwen2.5-Coder-32B-Instruct Test repair (3 rounds) 0%
QwQ-32B-Preview Single-shot 0%
QwQ-32B-Preview Compiler repair (3 rounds) 0%
QwQ-32B-Preview Test repair (3 rounds) 0%

2 models have held number one in the benchmark’s paper since its release in April 2025

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

OpenAI
Benchmark paper Not tested since May 2025
o3 48%

Source: The CRUST-bench 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%) o3 o1 o1-mini 0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) o3 o1

Source: Official CRUST-bench 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 18 months ago, in May 2025, for Claude Opus 4. 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 (12 entries).
  • Pass Models scored 7 models we track, plus 5 other entries on its leaderboard.
  • Warning Newest score The newest came 18 months ago, in May 2025, for Claude Opus 4.
  • Note Frontier models None of the top 10 has a score. The strongest tested is o3 (#84).
  • Pass Headroom The best is 48%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on arxiv.org: 22 rows for models we track and 5 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