Particle.news
Download on the App Store

Google Delays Gemini 3.5 Pro After Model Falls Short on Coding Benchmarks

Google is retuning the model with partners to meet engineers' coding standards under U.S. government review.

Overview

  • Bloomberg reported Thursday that Gemini 3.5 Pro is now months behind the June rollout Sundar Pichai announced at Google I/O.
  • Company sources say the flaghip missed internal performance goals on code generation and agentic coding, which are key for enterprise customers.
  • Google retrained Gemini with updated data in late June to improve coding ability but internal tests produced disappointing results.
  • The delay rattled markets, with Alphabet shares falling roughly 3–4% after the report, and has raised frustration among engineers who fear losing ground to rivals.
  • Rivals such as OpenAI and Anthropic have shipped stronger coding-focused models, and Google is consolidating internal teams and running partner and U.S. government tests to close the gap.