Tier 1 — floor
5/5
157 tokens · 4.691s (representative seed)
Tier 2 — restraint (refusal)
5/5
194 tokens · 1.533s (representative seed)
Tier 3 — distractor resistance
5/5
223 tokens · 1.705s (representative seed)
Tier 4 — multi-passage synthesis
0/5
1928 tokens · 13.848s (representative seed)
Tier 5 — precise extraction
5/5
281 tokens · 2.095s (representative seed)
Tier 6 — long-context needle
5/5
221 tokens · 1.779s (representative seed)