Benchmark Almanac

SCONE-bench

Claude Opus 4.5 $548,400,000 leads since December 2025
Not in model cards

Agents find and exploit vulnerabilities in real smart contracts on a local blockchain fork. Success is measured in simulated dollars stolen, not only pass or fail.

405 historical contracts at launch, later 417. A post-cutoff set of 12 incidents from 2026 was added in May 2026.

Released
December 2025
Kind
Finding and exploiting real bugs
Level
Expert
Scored
Checked automatically
Answers
Agent task
Tasks
All public
Strongest model tested
Claude Opus 4.5 #61 on the capability index
Last model added
November 2025
Scores refreshed
November 12th, 2026
9
Entries on its own leaderboard, 7 models we track

Claude Opus 4.5 leads with $548,400,000 among models released by November 2025

Best score per entry from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders.

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

  1. $548,400,000 Claude Opus 4.5
  2. $436,700,000 Claude Sonnet 4.5
  3. $410,200,000 Claude Opus 4.1
  4. $393,900,000 GPT-5
  5. $331,500,000 Claude Sonnet 4
  6. $12,000,000 o3
  7. $3,500,000 Claude 3.7 Sonnet
  8. $244,000 DeepSeek V3 (snapshot unspecified)
  9. $6,000 GPT-4o (snapshot unspecified)
  1. 1 Claude Opus 4.5 Best@8; original 405-contract full benchmark $548,400,000
  2. 2 Claude Sonnet 4.5 Best@8; original 405-contract full benchmark $436,700,000
  3. 3 Claude Opus 4.1 Best@8; original 405-contract full benchmark $410,200,000
  4. 4 GPT-5 Best@8; original 405-contract full benchmark $393,900,000
  5. 5 Claude Sonnet 4 Best@8; original 405-contract full benchmark $331,500,000
  6. 6 o3 Best@8; original 405-contract full benchmark $12,000,000
  7. 7 Claude 3.7 Sonnet Best@8; original 405-contract full benchmark $3,500,000
  8. 8 DeepSeek V3 (snapshot unspecified) Best@8; original 405-contract full benchmark $244,000
  9. 9 GPT-4o (snapshot unspecified) Best@8; original 405-contract full benchmark $6,000

Source: Official SCONE-bench leaderboard

#EntryScoreSettingDateSource
1 Claude Opus 4.5 $548,400,000 Best@8; original 405-contract full benchmark Nov 2025 Official leaderboard
2 Claude Sonnet 4.5 $436,700,000 Best@8; original 405-contract full benchmark Sep 2025 Official leaderboard
3 Claude Opus 4.1 $410,200,000 Best@8; original 405-contract full benchmark Aug 2025 Official leaderboard
4 GPT-5 $393,900,000 Best@8; original 405-contract full benchmark Aug 2025 Official leaderboard
5 Claude Sonnet 4 $331,500,000 Best@8; original 405-contract full benchmark May 2025 Official leaderboard
6 o3 $12,000,000 Best@8; original 405-contract full benchmark Apr 2025 Official leaderboard
7 Claude 3.7 Sonnet $3,500,000 Best@8; original 405-contract full benchmark Feb 2025 Official leaderboard
8 DeepSeek V3 (snapshot unspecified) $244,000 Best@8; original 405-contract full benchmark Official leaderboard
9 GPT-4o (snapshot unspecified) $6,000 Best@8; original 405-contract full benchmark Official leaderboard

Number one over time

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

Anthropic
Own leaderboard Not tested since Dec 2025
Claude Opus 4.5 $548,400,000

Source: Official SCONE-bench leaderboard

Scores over time

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
$0 $200,000,000 $400,000,000 $600,000,000 2025 2026 $0 $200,000,000 $400,000,000 $600,000,000 2025 2026

Source: Official SCONE-bench leaderboard

No lab model card reports it yet

We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark.

More in Cybersecurity

All 115 in Cybersecurity
  • ExploitGym In use 869 real-vulnerability exploit-development tasks (userspace, V8, Linux kernel). Success means capturing a flag through the intended vulnerability. GPT-6 Astra 42.4%
  • ExploitBench In use 41 recent V8 JavaScript-engine N-day vulnerabilities. Progress measured by 16 grader-verified capability flags across five tiers. GPT-6 Astra 100%
  • CyberBench v1.1 (Vals) In use Agents work on OSS-Fuzz memory-safety bugs from ARVO images in two tracks. They write a proof-of-concept input that crashes the target, or patch the source so the crash stops and tests still pass. Muse Wonder 80.4%
  • SRE-Bench In use Software reverse engineering without source: 262 binary instances from 19 private programs, each with six graded objectives. GPT-6 Astra 56.9%
  • CyScenarioBench In use Scenario-based long-horizon security challenges in sandboxes, organized in task, scenario and campaign tiers. Evaluation set kept private. Gemini 4 Argon 100%
  • ExploitBench - Internal Port In use Agents develop working exploits for recently disclosed vulnerabilities to limit contamination. Scored as arbitrary code execution success rate. GPT-6 Cyber 42%