BREAKING
Gemini 3.5 Pro Leak: 2M Tokens
0M
Gemini 3.5 Pro tokens
0M
Gemini 1.5 Pro window
0K
rumored output limit
Enthusiasts vs Skeptics
Enthusiasts
Load whole codebases
Less reliance on RAG
Lean on in-context learning
SkepticsMirage
Length isn't usable context
Higher latency and cost
Quality can degrade
Dump-Everything Trade-offs
Latency3
Cost8
Baseline1
Leak Not Yet Fact
1Leaked dev guides
2No Google confirmation
3No official benchmarks
4Vertex AI preview path
Expectation, Not Verified
AI NEWS BLITZ
Leaks claim Google's next model will carry a two-million-token context window.