Benchmark Almanac

VeruSAGE-Bench

Not in model cards
Version repository

Agents synthesize Verus proofs for 849 tasks extracted from eight real-world Rust systems projects.

The official leaderboard reports verified solve rates, time and cost for repository-level tasks.

Released
December 2025
Kind
Fixing real repositories
Level
Research frontier
Scored
Checked automatically
Answers
Code
Tasks
All public
Measures
Capability
Language
English
In and out
Text, Code → Code
Interaction
Single turn
Environment
Code sandbox
Items from
Real work and use
Refresh
One fixed set
Run by
Anyone
Based on
Verus-Bench
10
Entries on its own leaderboard, 0 models we track

VeruSAGE (Hands-off) + Claude Opus 4.5 tops the leaderboard with 93.9% No scores yet

Best score per entry from the benchmark's own leaderboard, read on November 21st, 2026. Where a model ran on its own with several settings, we keep its best. Every run is in the next chart. Agents other companies built around a model get a pale shade of that model's lab color. Gray marks blends and entries not in our model list, such as fine-tunes and older models. None of these count toward leaders. Agents built around a model, such as VeruSAGE (Hands-off) + Claude Opus 4.5, are their own entries. We don't count them as the model's score. This view shows only models we track, each at its best.

  1. 1 VeruSAGE (Hands-off) + Claude Opus 4.5 Hands-off 93.9%
  2. 2 VeruSAGE (Hands-off) + Claude Sonnet 4.5 Hands-off 80.9%
  3. 3 VeruSAGE (Hands-on) + Claude Sonnet 4.5 Hands-on 67.5%
  4. 4 VeruSAGE (Hands-off) + Claude Sonnet 4 Hands-off 64.1%
  5. 5 VeruSAGE (Hands-on) + Claude Sonnet 4 Hands-on 57.8%
  6. 6 VeruSAGE (Hands-on) + GPT-5 Hands-on 56.4%
  7. 7 VeruSAGE (Hands-off) + GPT-5 Hands-off 50.1%
  8. 8 VeruSAGE (Hands-on) + o4-mini Hands-on 41.1%
  9. 9 AutoVerus (Baseline) + o4-mini AutoVerus (Baseline) 19.8%
  10. 10 VeruSAGE (Hands-off) + o4-mini Hands-off 17%

Source: Official VeruSAGE-Bench leaderboard

#EntryScoreSettingReleasedRun onLinkSource
1 VeruSAGE (Hands-off) + Claude Opus 4.5 93.9% Hands-off Nov 2025 Jan 2026 arxiv.org Official leaderboard
2 VeruSAGE (Hands-off) + Claude Sonnet 4.5 80.9% Hands-off Sep 2025 Jan 2026 arxiv.org Official leaderboard
3 VeruSAGE (Hands-on) + Claude Sonnet 4.5 67.5% Hands-on Sep 2025 Jan 2026 arxiv.org Official leaderboard
4 VeruSAGE (Hands-off) + Claude Sonnet 4 64.1% Hands-off May 2025 Jan 2026 arxiv.org Official leaderboard
5 VeruSAGE (Hands-on) + Claude Sonnet 4 57.8% Hands-on May 2025 Jan 2026 arxiv.org Official leaderboard
6 VeruSAGE (Hands-on) + GPT-5 56.4% Hands-on Aug 2025 Jan 2026 arxiv.org Official leaderboard
7 VeruSAGE (Hands-off) + GPT-5 50.1% Hands-off Aug 2025 Jan 2026 arxiv.org Official leaderboard
8 VeruSAGE (Hands-on) + o4-mini 41.1% Hands-on Apr 2025 Jan 2026 arxiv.org Official leaderboard
9 AutoVerus (Baseline) + o4-mini 19.8% AutoVerus (Baseline) Apr 2025 Jan 2026 arxiv.org Official leaderboard
10 VeruSAGE (Hands-off) + o4-mini 17% Hands-off Apr 2025 Jan 2026 arxiv.org Official leaderboard
#ModelScoreSettingReleasedRun onLinkSource

The best score, 93.9%, costs $14 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.

Agents other companies built, in a pale shade of the model’s lab 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 VeruSAGE (Hands-off) + Claude Opus 4.5, Hands-off VeruSAGE (Hands-off) + Claude Sonnet 4.5, Hands-off VeruSAGE (Hands-on) + Claude Sonnet 4.5, Hands-on VeruSAGE (Hands-on) + GPT-5, Hands-on 0% 20% 40% 60% 80% 100% $0.10 $1.00 $10 $100 Cost per task, log scale VeruSAGE (Hands-off) + Claude Opus 4.5, Hands-off VeruSAGE (Hands-off) + Claude Sonnet 4.5, Hands-off

Source: Official VeruSAGE-Bench leaderboard

ModelEffortHarnessScoreCost per taskOutput tokens
VeruSAGE (Hands-off) + Claude Opus 4.5 Hands-off VeruSAGE (Hands-off) 93.9% $14
VeruSAGE (Hands-off) + Claude Sonnet 4.5 Hands-off VeruSAGE (Hands-off) 80.9% $5.61
VeruSAGE (Hands-on) + Claude Sonnet 4.5 Hands-on VeruSAGE (Hands-on) 67.5% $1.47
VeruSAGE (Hands-off) + Claude Sonnet 4 Hands-off VeruSAGE (Hands-off) 64.1% $6.65
VeruSAGE (Hands-on) + Claude Sonnet 4 Hands-on VeruSAGE (Hands-on) 57.8% $1.72
VeruSAGE (Hands-on) + GPT-5 Hands-on VeruSAGE (Hands-on) 56.4% $0.52
VeruSAGE (Hands-off) + GPT-5 Hands-off VeruSAGE (Hands-off) 50.1% $1.30
VeruSAGE (Hands-on) + o4-mini Hands-on VeruSAGE (Hands-on) 41.1% $0.67
AutoVerus (Baseline) + o4-mini AutoVerus (Baseline) AutoVerus (Baseline) 19.8% $0.74
VeruSAGE (Hands-off) + o4-mini Hands-off VeruSAGE (Hands-off) 17% $0.81

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.

Versions

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

All 8 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access All tasks are public.
  • Pass Confidence Checked against a primary source.
Scores 2 pass 3 notes
  • Pass Independent runs Scores from the official leaderboard (10 entries).
  • Pass Models scored 10 entries on its leaderboard, none a model we track.
  • Note Newest score No scores yet.
  • Note Frontier models No models scored yet.
  • Note Headroom No scores yet.
Leaderboard import 3 pass
  • Pass Importer We read the board on microsoft.github.io: 10 rows for models we track at the last read.
  • Pass Last read November 21st, 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