Gemini 3.6 Flash Leak — ID Spotted, First Outputs Rough
· 5 hours ago
A new model ID — "gemini-3.6-flash-tiered" — just surfaced in Antigravity, and someone already ran first tests.
The early outputs are rough, per the tester. That's normal, leaked checkpoints are unfinished. Judge it when it ships, not when it leaks.
What we know so far Gemini 3.6 Flash Leak — First Tests
More from the feed
Tip
Tip, Keep a Local Copy of the Weights You Depend On
Quick tip, given the news, if a Chinese open-weight model matters to your workflow, keep a local copy of the weights.
Hosted access can change — a policy, an API shutdown, a region block. Local weights can't be revoked. This is prudent, not panic.
Why this is suddenly relevant White House Weighs Executive Order on Chinese Open-Weight AI
New model
Qwen3.8-Max-Preview Keeps Improving Daily
Qwen3.8-Max-Preview is shipping improvements daily, and the web-frontend gains are the ones you'll actually notice.
The best part, the official version will be open-weight. What you test now, you can self-host later.
The daily progress log Qwen3.8-Max-Preview Improves Daily
New model
Google Frozen v2 — Gemini Architecture Baked Into Silicon
Google is reportedly working on Frozen v2, a chip with the Gemini architecture baked directly into silicon. The claim is 6–10x better tokens per watt, targeting 2028.
The trade-off is rigidity — hardwire the architecture and you bet it stays relevant. Bold bet.
Full report Google Frozen v2 — The Gemini Chip
