Google DeepMindPreview
Gemini 3.5 Pro
Google’s top reasoning tier — announced, not yet shipped. Verdict pending GA.
Read the full Gemini 3.5 Pro analysisContext
—
Max output
—
Input /1M
—
Output /1M
—
Best for
- Hard reasoning where Flash hits its ceiling (once GA)
- Workloads needing 2M context across the widest modality set (targeted)
- Heavy video/audio multimodal reasoning
Watch out
Still in limited Vertex preview as of June 5, 2026 — no model card, no benchmarks, no pricing. GA targeted for June. Build on Gemini 3.5 Flash in the meantime.
For creators. The one to watch for agentic video/multimodal pipelines once GA confirms Deep Think numbers. For now, route creator agent work to Gemini 3.5 Flash.
Capabilities
- Deep Think reasoning (vendor-stated)
- Targeted 2M-token context window (vendor-stated)
- Frontier multimodal: text/image/audio/video/PDF
- Positioned as Google's strongest agentic + coding model (vendor-stated)