Google Rumored to Shelve Gemini 3.5 Pro, Pivot to Gemini 4

Google’s long-awaited flagship AI model, Gemini 3.5 Pro, may never arrive. A report from research firm SemiAnalysis claims Google has shelved the repeatedly delayed model and is redirecting core engineering resources toward the next-generation Gemini 4, with the recently deployed Gemini 3.6 Flash serving as a stopgap. The report paints a blunt picture of Google DeepMind’s standing in the frontier model race.

Gemini 3.5 Pro reportedly shelved

According to the SemiAnalysis report, Google has officially shelved Gemini 3.5 Pro after months of delays. Rather than push through the deployment bottlenecks and performance hurdles that stalled the flagship model, Google is reportedly moving its engineering resources to Gemini 4 and leaving Gemini 3.6 Flash as a high-speed stopgap.

Rankings and competitive pressure

The report states that internal testing and external benchmarks placed Gemini models around 8th or 9th overall in frontier capability rankings, trailing rivals including Muse Spark 1.2, Grok 4.5, and open-source models that have rapidly closed the reasoning and coding gap. Lighter Gemini models such as Gemini 3.5 Flash and 3.6 Flash have earned praise for inference speed and token efficiency in agentic workflows, but the heavier Pro tier has faced persistent friction from deployment glitches, edge-case reasoning errors, and infrastructure bottlenecks.

Google pushes back

Google spokesperson Logan Kilpatrick dismissed the SemiAnalysis breakdown as “superficial,” pointing to the technical progress across the ecosystem and the rapid rollout of Gemini 3.6 Flash as evidence of strong execution. Pre-training for Gemini 4 is reportedly already underway on Google’s custom TPU clusters, signaling a plan to leapfrog current limitations rather than fight incremental battles.

SDK hints at a “Gemini 3.5 Pro+”

Separately, developer Choblin spotted fresh commits in Google’s official GenAI SDK repository referencing a “Gemini 3.5 Pro+” model. The updated logic shows that while legacy models use fixed-rate frame sampling for video processing, Gemini 3.5 Pro+ defaults to a new AGENTIC mode driven by “model-driven dynamic navigation.” Instead of passively ingesting uniformly sampled frames, the model decides where and when to look, skipping low-information scenes and preserving context window for higher-value reasoning.

FAQ

Has Gemini 3.5 Pro been officially cancelled?

No. The SemiAnalysis report is unconfirmed, and Google has pushed back on the claims without issuing a formal statement on the model’s status.

What is the difference between Gemini 3.5 Pro and Pro+?

Per the SDK code, Pro+ defaults to an AGENTIC mode with model-driven dynamic navigation for video, replacing fixed-rate frame extraction.

When will Gemini 4 launch?

Google has not announced a date. Reports only indicate that pre-training is underway on Google’s TPU clusters.

📌 Reminder: the details above are leaks and rumors that Google has not confirmed. Model names, status, and specifications may differ from the final release.

Cover image generated by AI for illustrative purposes only.

Sources: NokiaPowerUser · Geeky Gadgets