← Back to leaderboard

llama3.2:3b

3.1 GB measured VRAM · sub-4GB tier · rag-gauntlet-v1
Exact passages, questions, and expected answers for this category are held back from publication to avoid training-data contamination — see About for why. This page shows per-tier pass/fail counts and timing only.
Tier 1 — floor 5/5
65 tokens · 2.313s (representative seed)
Tier 2 — restraint (refusal) 5/5
47 tokens · 0.353s (representative seed)
Tier 3 — distractor resistance 5/5
35 tokens · 0.318s (representative seed)
Tier 4 — multi-passage synthesis 0/5
98 tokens · 0.555s (representative seed)
Tier 5 — precise extraction 5/5
37 tokens · 0.309s (representative seed)
Tier 6 — long-context needle 5/5
17 tokens · 0.273s (representative seed)