← Claude Sonnet 5 · API · High effort (default)
Evidenceacceptedthird_party_leaderboard
Chatbot Arena / LMArena human preference
Chatbot Arena leaderboard mirror (openlm.ai) — Claude Sonnet 5 Thinking
ev_2026_07_sonnet5_arena
Result
1485
Confidence B
- Tested
- 2026-07-11
- Runs
- 1
Configuration
How it was measured
- Exact model
- claude-sonnet-5
- Reasoning
- high
- Agent harness
- none
- Tool access
- chat product surface (not tool-agent harness)
- Prompting
- anonymous pairwise battles (Arena)
PRIMARY independent CHA signal — blind human preference Elo on Chatbot Arena style leaderboard (openlm.ai mirror snapshot).
Caveats
Known limitations
- Preference Elo ≠ precise constraint following (pair with IF when available).
- Product chat surface may differ from API high-effort coding configs.
- Mirror site; confirm against primary LMArena when publishing.
Weighting
Evidence weight factors
The engine weights each evidence row by these factors (0–1) when fusing scores — higher is more trusted.
relevance0.90
quality0.85
recency1.00
health0.85
comparability0.75
reliability0.80
Benchmark
Chatbot Arena / LMArena human preference
Blind pairwise human preference battles aggregated into Elo ratings for chat quality, style, and helpfulness under real user prompts.
Benchmark source →