UK AISI and CAISI simulated enterprise network range
Agents attack a 32-step simulated enterprise network that takes a human expert about 20 hours. The score is steps completed.
Government pre-deployment evaluation. AISI names this range The Last Ones. The range is not public. Kimi K3 report cites the joint assessment.
- Made by
- UK AI Security Institute and NIST CAISI
- Kind
- Capture-the-flag challenges
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- Kept private
- Strongest model tested
- Kimi K3 #24 on the capability index
- Last in a system card
- July 2026
- Last model added
- July 2026
- Scores refreshed
- November 12th, 2026
Claude Opus 4.6 leads with 15.6 among models released by February 2026
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. 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.
Kimi K3 leads with 17.0, claims included
Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model.
The newest model with a score here came out in February 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 16 more
- 1 Claude Opus 4.6 100M token budget 15.6
- 2 Claude Opus 4.5 100M token budget 11.0
- 3 GPT-5.3 Codex 100M token budget 11.0
- 4 GPT-5.1-Codex 100M token budget 9.80
- 5 Claude Sonnet 4.5 100M token budget 9.40
- 6 Claude Sonnet 3.7 10M token budget 5.80
- 7 GPT-4o Aug 2024, 10M token budget 1.70
Source: Official UK AISI and CAISI simulated enterprise network range leaderboard
- 1 Kimi K3 17.0
- 2 Claude Opus 4.6 100M token budget 15.6
- 3 Claude Opus 4.5 100M token budget 11.0
- 4 GPT-5.3 Codex 100M token budget 11.0
- 5 GPT-5.1-Codex 100M token budget 9.80
- 6 Claude Sonnet 4.5 100M token budget 9.40
- 7 GPT-4o Aug 2024, 10M token budget 1.70
Sources: Labs’ model cards and official UK AISI and CAISI simulated enterprise network range leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 4.6 | 15.6 | 100M token budget | Feb 2026 | Official leaderboard |
| 2 | Claude Opus 4.5 | 11.0 | 100M token budget | Nov 2025 | Official leaderboard |
| 3 | GPT-5.3 Codex | 11.0 | 100M token budget | Feb 2026 | Official leaderboard |
| 4 | GPT-5.1-Codex | 9.80 | 100M token budget | Nov 2025 | Official leaderboard |
| 5 | Claude Sonnet 4.5 | 9.40 | 100M token budget | Sep 2025 | Official leaderboard |
| 6 | Claude Sonnet 3.7 | 5.80 | 10M token budget | Official leaderboard | |
| 7 | GPT-4o Aug 2024 | 1.70 | 10M token budget | Aug 2024 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Kimi K3 | 17.0 | Jul 2026 | Lab model card | |
| 2 | Claude Opus 4.6 | 15.6 | 100M token budget | Feb 2026 | Official leaderboard |
| 3 | Claude Opus 4.5 | 11.0 | 100M token budget | Nov 2025 | Official leaderboard |
| 4 | GPT-5.3 Codex | 11.0 | 100M token budget | Feb 2026 | Official leaderboard |
| 5 | GPT-5.1-Codex | 9.80 | 100M token budget | Nov 2025 | Official leaderboard |
| 6 | Claude Sonnet 4.5 | 9.40 | 100M token budget | Sep 2025 | Official leaderboard |
| 7 | GPT-4o Aug 2024 | 1.70 | 10M token budget | Aug 2024 | Official leaderboard |
Labs report scores for 1 model in their cards. Switch the scores above to rank them together.
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.
Source: Official UK AISI and CAISI simulated enterprise network range leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Claude Opus 4.6 | 100M token budget | 15.6 | |||
| Claude Opus 4.5 | 100M token budget | 11.0 | |||
| GPT-5.3 Codex | 100M token budget | 11.0 | |||
| GPT-5.1-Codex | 100M token budget | 9.80 | |||
| Claude Opus 4.6 | 10M token budget | 9.80 | |||
| Claude Sonnet 4.5 | 100M token budget | 9.40 | |||
| GPT-5.1-Codex | 10M token budget | 8.00 | |||
| Claude Opus 4.5 | 10M token budget | 7.60 | |||
| GPT-5.3 Codex | 10M token budget | 7.20 | |||
| Claude Sonnet 4.5 | 10M token budget | 6.10 | |||
| GPT-4o Aug 2024 | 10M token budget | 1.70 |
5 models have held number one on the official leaderboard since 2023 6 models have held number one with claims included since 2023
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official UK AISI and CAISI simulated enterprise network range leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Sources: Official UK AISI and CAISI simulated enterprise network range leaderboard and labs’ model cards
Labs report it only in system cards
We read the headline results of 522 launch posts from 52 labs, and none of them list this benchmark. Some system cards do, below.
One system card reports it
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| Jul 27, 2026 | Kimi K3 technical report | 17 | 32-step simulated enterprise network, 17 steps completed, UK AISI and CAISI joint assessment cited in 6.2.2 | Cyber risk |
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%