Google Delays Gemini 3.5 Pro, Struggles to Improve Coding Performance
Key Takeaways
- ▸Gemini 3.5 Pro missed its June 2026 launch deadline due to inadequate coding performance despite recent training-data updates
- ▸Google is testing Gemini 3.5 Pro and an upgraded Flash model with partners before public release, with no revised timeline announced
- ▸Internally, 75% of Google's new code is now AI-generated, yet some engineering teams resist AI-written code on quality grounds
Summary
Google has missed its June launch deadline for Gemini 3.5 Pro, with the company publicly acknowledging it is 'currently testing' the model alongside an upgraded Flash version. According to Bloomberg, the delay stems from disappointing results in improving coding performance—a critical differentiator in competitive AI markets. Google updated its training data in late June to boost coding capabilities, but the effort failed to meet internal benchmarks.
The delay marks a notable setback in Google's AI roadmap, leaving Gemini 3.1 Pro (from February) as the latest major public release. Internally, Google faces compounding challenges: 75% of newly written code is now AI-generated and approved by engineers, yet some teams harbor skepticism about AI-generated code quality. Additionally, the company is consolidating fragmented AI coding tools across DeepMind (AI Studio), Cloud (Vertex), and Android teams—a sign of both expanding AI adoption and resource strain.
Google stated it remains committed to 'shipping quickly across a wide range of models while keeping them highly cost-effective,' signaling that cost and performance must both improve before Gemini 3.5 Pro reaches users. No revised launch timeline has been announced.
- Google is consolidating multiple AI coding tools (DeepMind, Cloud, Android) to address tool fragmentation and capacity constraints
