← DeepSeek V4 Pro · API · Max effort
Evidenceacceptedindependent_lab
Humanity's Last Exam
Artificial Analysis — Humanity's Last Exam (DeepSeek V4 Pro max)
ev_2026_07_dsv4pro_aa_hle
Result
35.9%
Confidence A
- Tested
- 2026-07-11
- Runs
- 1
Configuration
How it was measured
- Exact model
- deepseek-v4-pro
- Reasoning
- max
- Agent harness
- none
- Tool access
- none (closed-book)
- Prompting
- Artificial Analysis HLE protocol (AA suite)
Secondary INT — independent AA HLE (closed-book). Frontier-relative norm under meth_v0_1_1. Float locked via AA-sourced OpenRouter model card.
Caveats
Known limitations
- Closed-book AA protocol — not tool-enabled multi-agent HLE.
- Float transcribed from AA-sourced OpenRouter mirror of AA evaluation suite.
- Materially lower than some vendor tool-stack HLE figures if those appear later.
Weighting
Evidence weight factors
The engine weights each evidence row by these factors (0–1) when fusing scores — higher is more trusted.
relevance0.95
quality1.00
recency1.00
health0.90
comparability0.85
reliability0.85
Benchmark
Humanity's Last Exam
Frontier closed-ended academic benchmark across many subjects; often reported with and without tools.
Benchmark source →