BREAKING
Gemini 3.6 Flash & 3.5 Flash-Lite Go GA
Two Models, Two Niches
Gemini 3.6 Flash
Successor
●
Better token efficiency
●
Stronger coding & planning
●
Lower price point
3.5 Flash-Lite
Low latency
●
Cost-efficient option
●
Built for sub-agents
●
High-volume automation
Gemini 3.6 Flash Benchmarks
SWE-Bench Pro
58.7
Terminal-Bench 2.1
78
OSWorld-Verified
83
0
M
token context
0
x
faster output
0
$
per M input tokens
Sampling Params Deprecated
Cheaper, Faster Tiers Intensify
AI NEWS BLITZ
Google DeepMind moves two lightweight Gemini models to general availability.