BREAKING
Kimi K3 Gets 3 Reasoning Levels
Standard, High and Max Effort
1Standard
2High
3Max
0T
total parameters
0M
token context
Max-Effort Benchmarks
BrowseComp91.2
Terminal Bench 2.188.3
Praise Meets Cost Concerns
Strengths
Frontier-level reasoning
Strong coding, agentic tasks
Complaints
Slow and verbose
High real-world cost
Open Weights Due July 27
AI NEWS BLITZ
Moonshot AI now lets users choose how hard Kimi K3 thinks on web and mobile.