Agents · Benchmark profile
Professional-level Capture-the-Flag cybersecurity agent benchmark measuring unguided end-to-end task solve rates.
Data verified 15 July 2026 · Methodology 1.2.0
Benchmark score on Cybench
Claude Mythos 5 leads at 100%, followed by Claude Opus 4.7 (96%) and Claude Opus 4.6 (93%).
Visual analysis
Three views of the same sourced leaderboard: model placement, the score curve, and descriptive provider averages. These visuals never alter ranking calculations.
7 ranked models · higher is better
| Model | Rank | Provider | Score |
|---|---|---|---|
| Claude Mythos 5 | #1 | Anthropic | 100% |
| Claude Opus 4.7 | #2 | Anthropic | 96% |
| Claude Opus 4.6 | #3 | Anthropic | 93% |
| Muse Spark 1.1 | #4 | Meta | 92.9% |
| Claude Opus 4.5 | #5 | Anthropic | 82% |
| Muse Spark | #6 | Meta | 65.4% |
| Grok 4.1 | #7 | xAI | 39% |
Shows how quickly scores separate down the field
| Model | Rank | Provider | Score |
|---|---|---|---|
| Claude Mythos 5 | #1 | Anthropic | 100% |
| Claude Opus 4.7 | #2 | Anthropic | 96% |
| Claude Opus 4.6 | #3 | Anthropic | 93% |
| Muse Spark 1.1 | #4 | Meta | 92.9% |
| Claude Opus 4.5 | #5 | Anthropic | 82% |
| Muse Spark | #6 | Meta | 65.4% |
| Grok 4.1 | #7 | xAI | 39% |
Average within the visible cohort · descriptive only
| Provider | Models | Average score |
|---|---|---|
| Anthropic | 4 | 92.8% |
| Meta | 2 | 79.2% |
| xAI | 1 | 39% |
One best score per model · higher is better
| Rank | Model | Provider | License | Score |
|---|---|---|---|---|
| #1 | Claude Mythos 5 claude-mythos-5 | Anthropic | closed | 100% |
| #2 | Claude Opus 4.7 claude-opus-4-7 | Anthropic | closed | 96% |
| #3 | Claude Opus 4.6 claude-opus-4-6 | Anthropic | closed | 93% |
| #4 | Muse Spark 1.1 muse-spark-1-1 | Meta | closed | 92.9% |
| #5 | Claude Opus 4.5 claude-opus-4-5 | Anthropic | closed | 82% |
| #6 | Muse Spark muse-spark | Meta | closed | 65.4% |
| #7 | Grok 4.1 grok-4-1 | xAI | closed | 39% |
The top of this snapshot is led by Claude Mythos 5 at 100%; third place is 7.0 points behind. The top-7 spread is 61.0 points.
About Cybench
Professional-level Capture-the-Flag cybersecurity agent benchmark measuring unguided end-to-end task solve rates. Results stay tied to the exact model variant and evaluation system. Multiple systems for the same model use the best published score on this page; overall Lumina scoring uses the median of ranking-eligible rows.
Open benchmark source ↗FAQ
Professional-level Capture-the-Flag cybersecurity agent benchmark measuring unguided end-to-end task solve rates.
Claude Mythos 5 by Anthropic currently leads with 100%.
7 models in the LuminaBench cohort have a qualifying score on this benchmark.
Yes. This family is ranking-weighted at 1.0% of its capability category. Claude Mythos 5 is overall #1.
Related
