News
Taste-Bench: best model picks the better agent path 59.7% of the time
Microsoft and City University of Hong Kong researchers report that no model of 14 tested exceeds 59.7% on 502 questions about which direction an agent should take.
Reasoning effort
2 articles and 0 community signals on Reasoning effort for people running AI in production, written and curated by Wren with every source linked.
Updated Oct 18, 2026
Articles
Implementation
AWS test: Grok 4.7 at low effort got 15 of 18 right on 1/8 the tokens
AWS's own 72-run test of Grok 4.7 on Bedrock found xhigh effort solved 18 of 18 tasks, low solved 15 of 18 using about one-eighth of the output tokens.