Tier 1 — floor
5/5
16 tokens · 2.265s (representative seed)
Tier 2 — restraint (refusal)
5/5
35 tokens · 0.256s (representative seed)
Tier 3 — distractor resistance
5/5
24 tokens · 0.223s (representative seed)
Tier 4 — multi-passage synthesis
0/5
47 tokens · 0.372s (representative seed)
Tier 5 — precise extraction
5/5
46 tokens · 0.365s (representative seed)
Tier 6 — long-context needle
5/5
21 tokens · 0.322s (representative seed)