Rumor: Gemini 3.5 Pro Retraining Fails
APPSO · wechat · 2026-07-17
According to APPSO, citing internal sources and public reports like Bloomberg, Google's Gemini 3.5 Pro is facing significant hurdles:
- Code generation capabilities fell short of internal expectations. Despite a supplemental training data run in late June, scores showed no notable improvement.
- Rumor has it that Google has scrapped the original training plan and might retrain the model from scratch.
- Worse, internal evaluations indicate it might even lag behind Meta's MuseSpark. The model comparison chart previously accessible to employees has had the 3.5 Pro scores removed.
- Google has revoked internal testing access for staff. Morale in related teams is low, with some employees considering jumping ship to other AI labs.
The article also notes that talent drain and delayed releases related to Gemini have indirectly impacted Alphabet's market cap.
Related event: Gemini 3.5 Pro Delayed by Months Over Coding Performance(12 posts)→
More from Companies & People
- FactoryAI gave back its first millions, then shipped Droid CLI two years later — matanSF · 2026-07-22
- A post says AI teams should drop the research scientist vs engineer split — jsuarez · 2026-07-22
- YC startup Vendo launches an open-source customization layer for user-built micro-apps — ycombinator · 2026-07-22
- Prescience launches publicly as an AI-native health insurance company — ycombinator · 2026-07-22
- SF AI crowd swaps poker for a bullet and bughouse chess night — Jackyhuang · 2026-07-22
- Polymarket puts Anthropic’s year-end IPO odds at 64% amid patent suit — Polymarket · 2026-07-22