Gemini 3.5 Pro: Fourth Delay, No GA Date

Gemini 3.5 Pro — Google's would-be third US frontier model — is reported to be on its fourth postponement and "months behind schedule," Bloomberg reported July 16, citing 10 current and former Google employees. The core shortfall: coding capabilities that fall short of internal benchmarks. The model announced at Google I/O in May with a June launch target now has no confirmed GA date.

What happened

Google updated Gemini 3.5 Pro's training data late last month in an attempt to close the performance gap — and the results missed internal targets. The reported delay timeline has stretched from June → late June → July → indefinite, according to Bloomberg.

In response, Google is reportedly preparing a stopgap "Gemini 3.6 Flash" while 3.5 Pro remains in the lab, Bloomberg reports. A Google spokesperson told the outlet: "We're shipping quickly across a wide range of models while keeping them highly cost-effective for customers." The models shipping are Flash variants — not the Pro frontier model announced at I/O.

Why Gemini 3.5 Pro's delay matters

The US frontier gap is widening. Only two US frontier models are shipping today — Claude Fable 5 and GPT-5.6 Sol — while three Chinese open-weight frontier models launched this month: Tencent Hy3, Z.ai GLM-5.2, and Moonshot AI's Kimi K3, a 2.8T-parameter model that reportedly beat Fable 5 for #1 on Arena Code WebDev the same day Bloomberg published the delay.

Alphabet shareholders are paying attention. The stock dropped ~4% (Bloomberg reports roughly $225 billion in market cap was erased in a single session). Alphabet has guided for $180–190 billion in full-year 2026 capex — more than double 2025 spend — and expects that figure to rise again in 2027 (Bloomberg; CNBC, 2026).

The brain drain compounds the problem. Four senior DeepMind researchers departed in a single week in late June: Noam Shazeer (Gemini co-lead) to OpenAI, and John Jumper, Jonas Adler, and Alexander Pritzel to Anthropic. DeepMind staff have internally flagged that Google lacks a credible commercial coding product — from the company that invented the Transformer architecture (Bloomberg, 2026).

What changes for you

If you were waiting for Gemini 3.5 Pro as a third frontier option, stop. Google has not committed to a new timeline, and the "3.6 Flash" stopgap is not a frontier model.

• Evaluate Claude Sonnet 5 ($3/$15 per 1M tokens) or GPT-5.6 Sol ($5/$30) for production deployments. Both are shipping, independently benchmarked, and GA-priced with stable APIs.
• Monitor the open-weight frontier. Kimi K3 is now competitive with US frontier models on coding benchmarks — it may serve as a third option without waiting for Google.
• Watch for Gemini 3.6 Flash in Q3. It won't match frontier coding, but for multimodal and 1M+ token context workloads, it may be a viable stopgap until 3.5 Pro ships.

FAQ

Will Gemini 3.5 Pro ship in 2026? Unknown. Google has not provided a revised timeline. The reported fourth delay and "months behind" language from Bloomberg suggest Q4 at the earliest, but the model could slip further or ship below internal targets.

What is Gemini 3.6 Flash? A stopgap model reportedly in preparation while 3.5 Pro remains in training. Flash variants typically prioritize speed and cost over raw capability. Do not expect frontier-level coding from a Flash release.

How does this affect Google's frontier position? Gemini 3.5 Pro's 2M-token context window, native multimodal pipeline, and Deep Think reasoning remain real capabilities — if and when they ship. But the competitive window is closing. Every month without a shipping Pro model widens the gap Google has to close, and the open-weight frontier models are not waiting.