← Claude Sonnet 5 · API · High effort (default)
Evidenceacceptedindependent_lab
Humanity's Last Exam
Artificial Analysis — Humanity's Last Exam (Claude Sonnet 5 max)
ev_2026_07_sonnet5_aa_hle
Result
39.57%
Confidence A
- Tested
- 2026-07-01
- Runs
- 1
Configuration
How it was measured
- Exact model
- claude-sonnet-5
- Reasoning
- max
- Agent harness
- none
- Tool access
- AA default (do not assume Anthropic system-card tool stack)
- Prompting
- Artificial Analysis HLE protocol (AA suite)
Independent HLE from Artificial Analysis. NOT comparable to vendor HLE-with-tools 57.4% (ev_2026_07_sonnet5_hle) — AA protocol yields ~39.6% for Sonnet 5 max. normalisedScore uses meth_v0_1_1 frontier-relative HLE map (anchor 55% → 100).
Caveats
Known limitations
- Materially lower than vendor tool-enabled HLE; always label tool policy.
- Prefer this row for independent closed-ended hard reasoning comparisons.
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 →