Google delays Gemini 3.5 Pro as model lags rivals on coding; Alphabet shares slip

According to a Bloomberg report picked up by Reuters and MarketWatch, Google's next flagship, Gemini 3.5 Pro, has slipped from an original June window to an unconfirmed July 17 target after failing to clear internal benchmarks, especially on coding tasks. Google engineers retrained the model on updated data in late June, but the results reportedly still came in below goals, and the model remains missing from public API documentation. Alphabet's stock fell on the news, with analysts framing the delay as evidence Google is struggling to keep pace in the agentic-coding race.

The competitive backdrop is what makes the slip sting. Over the same window, rivals have shipped coding-focused agentic models: OpenAI's GPT-5.6 (Sol/Terra/Luna), xAI's Grok 4.5 — its 'first model specifically trained for coding and agents' — and Moonshot's open-weight Kimi-K3 all landed first. Engineers inside Google are reportedly anxious about the pace, and community sentiment has curdled: threads on X and Reddit repeatedly cite the June→July→'unconfirmed' pattern as a confidence problem.

Google is trying to change the subject with product news, rebranding NotebookLM as Gemini Notebook and promising Gemini 3.5 and Antigravity upgrades to AI Pro subscribers and Search's AI Mode. But until a shipping model with named benchmarks appears, the narrative is a leader visibly falling behind. Watch for whether Google ships anything on the July 17 target date or lets it slip again — a second miss would harden the 'Google can't keep up on coding' storyline that rivals are happy to amplify.