← Gemini 3.1 Pro · API · High thinking
Evidenceacceptedvendor_reported
Humanity's Last Exam
Google DeepMind — Gemini 3.1 Pro · HLE (no tools)
ev_2026_07_gem31pro_hle_vendor
Result
44.4%
Confidence B
- Tested
- 2026-02-19
- Runs
- 1
Configuration
How it was measured
- Exact model
- gemini-3.1-pro-preview
- Reasoning
- high
- Agent harness
- none
- Tool access
- none (closed-book)
- Prompting
- vendor HLE
Secondary INT — vendor closed-book HLE. Prefer AA float when locked; frontier-relative norm under meth_v0_1_1.
Caveats
Known limitations
- Vendor-reported; cannot receive confidence S.
- Tool-enabled HLE 51.4% is a different protocol — not this row.
- Upgrade when independent AA HLE float is locked.
Weighting
Evidence weight factors
The engine weights each evidence row by these factors (0–1) when fusing scores — higher is more trusted.
relevance0.95
quality0.55
recency0.85
health0.90
comparability0.85
reliability0.70
Benchmark
Humanity's Last Exam
Frontier closed-ended academic benchmark across many subjects; often reported with and without tools.
Benchmark source →