Tier 1 — floor
5/5
14 tokens · 9.955s (representative seed)
Tier 2 — restraint (refusal)
5/5
17 tokens · 0.5s (representative seed)
Tier 3 — distractor resistance
5/5
21 tokens · 0.616s (representative seed)
Tier 4 — multi-passage synthesis
5/5
268 tokens · 4.848s (representative seed)
Tier 5 — precise extraction
5/5
47 tokens · 1.038s (representative seed)
Tier 6 — long-context needle
5/5
20 tokens · 0.848s (representative seed)