Last Secure Code Benchmark
Agents rebuild code that a real vulnerability once affected, from a specification that never mentions security. Hidden tests check function and replay an exploit.
450 tasks from 150 vulnerability scenarios, 115 from real CVEs and 35 expert written. Seven languages and 24 CWE classes. Each scenario appears at function, file and repository level. The headline metric counts a task only when both the function and security suites pass. Apache 2.0, also on Harbor Hub and Docker Hub.
- Released
- October 2026
- Kind
- Secure code and review
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Code
- Tasks
- All public
- Strongest model tested
- Claude Fable 5 #11 on the capability index
- Last model added
- August 2026
- Scores refreshed
- November 12th, 2026
In the authors’ words
AI coding agents now write much of the code that ships, yet whether that code is secure remains unclear. We introduce the Last Secure Code Benchmark, a comprehensive evaluation of secure code generation by coding agents. Built from 150 real vulnerabilities, its 450 tasks span seven widely used languages (Java, C, C++, JavaScript, Python, PHP, and Go), 24 CWEs including 21 of the 2025 CWE Top 25 and 8 of the Top 10 KEV weaknesses, more than any benchmark we compare with, and three vibe-coding scenarios: filling in a function, writing a file, and building a repository's source tree from scratch. Every task asks only for functionality, through a specification that never mentions security, and hidden functional and security tests grade the result. Across six frontier agents, 73.6% to 91.3% of tasks pass functionally, yet even the strongest, Claude Fable 5, produces code that is both functional and secure for only 25.1% of tasks, and the others for 16.4% to 24.7%. The best model only passes 12% of the vulnerability scenarios across all vibe coding setting, posing strong concern on the security of the code written by vibe coding.
Claude Fable 5 leads with 25.1% among models released by August 2026
Best score per model from the benchmark's own leaderboard, read on November 12th, 2026. Where a model ran with several settings or harnesses, we keep its best.
The newest model with a score here came out in August 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- Claude Fable 5.2
-
Gemini 4 Carbon - GPT-6 Astra
- Claude Opus 5.5
- and 14 more
- 1 Claude Fable 5 25.1%
- 2 GPT-5.6 Sol 24.7%
- 3
Qwen 3.8 Max 22% - 4 GLM-5.3 21.3%
- 5
DeepSeek-V4-Pro 18% - 6 Kimi K3 16.4%
Source: Official Last Secure Code Benchmark leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Fable 5 | 25.1% | Claude Code | Jun 2026 | Official leaderboard |
| 2 | GPT-5.6 Sol | 24.7% | Codex | Jul 2026 | Official leaderboard |
| 3 | Qwen 3.8 Max | 22% | Claude Code | Aug 2026 | Official leaderboard |
| 4 | GLM-5.3 | 21.3% | Claude Code | Aug 2026 | Official leaderboard |
| 5 | DeepSeek-V4-Pro | 18% | Claude Code | Apr 2026 | Official leaderboard |
| 6 | Kimi K3 | 16.4% | Claude Code | Jul 2026 | Official leaderboard |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official Last Secure Code Benchmark leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official Last Secure Code Benchmark 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.
Versions
More 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%