squad: 0
est cost/run: $0
⏱ —
No race squad selected.
Pick models in the sidebar → RACE SQUAD, then LAUNCH.
Tip: TOP 12 preloads the frontier. You can also hit the quick chips above the model list.
Pick models in the sidebar → RACE SQUAD, then LAUNCH.
Tip: TOP 12 preloads the frontier. You can also hit the quick chips above the model list.
selected: 0
est cost/run: $0
Battle Royale = fire the same prompt at dozens of models.
Pick models on the left, set concurrency, DEPLOY. Results stream in below with cost / latency / verdict.
Pick models on the left, set concurrency, DEPLOY. Results stream in below with cost / latency / verdict.
1 · TARGET PROMPT
2 · MUTATIONS 0 selected
3 · PREVIEW (applied in order)
select mutations above…
4 · TARGET MODEL
ASSAULT MATRIX fires the original + every selected mutation at the target model in parallel,
then scores each response: did it comply, refuse, or get confused?
Use on your own systems to harden them — not on things you don't own.
Mutations prepared. Run an ASSAULT MATRIX to see robustness scores per variant.
Compliance % = how eagerly the model followed the obfuscated instruction.
Compliance % = how eagerly the model followed the obfuscated instruction.