$ cat wiki/models/gemini-4.md
Gemini 4
modelupdated 2026-07-28created 2026-07-28
Spec
| Attribute | Value |
|---|---|
| Developer | Google DeepMind |
| Released | not yet |
| Announced | 2026-07-21 (pre-training confirmation buried in Gemini 3.6 Flash post) |
| Context window | unknown |
| Pricing | unknown |
| License | proprietary |
| Availability | unknown |
Release Date
- Pre-training confirmed: July 21, 2026 (Sundar Pichai, buried in Gemini 3.6 Flash announcement)
- Pre-training status: actively running as of July 2026
- Expected release: unknown (no target announced)
Key Claims (as of July 21, 2026)
- Sundar Pichai: "most ambitious pre-training run yet"
- "Significantly larger than any prior Gemini model"
- Primary capability targets: coding and autonomous agents
- Pre-training running in parallel with Gemini 3.5 Pro enterprise preview (not a replacement)
⚠️ No benchmarks, parameters, pricing, context window, or release timeline have been disclosed.
Context
The confirmation came as a secondary disclosure inside the Gemini 3.6 Flash launch post rather than a standalone announcement — hence its late appearance in this wiki.
Gemini 3.5 Pro (the current un-released model in enterprise preview) had already missed its June and July 17 GA targets; Gemini 4 is a longer-horizon successor.
Related
- Gemini 3.6 Flash — current GA release (Jul 21 2026)
- Gemini 3.5 Pro — still in enterprise preview, GA delayed
- Google DeepMind — originating lab
- Claude Fable 5 — current frontier model it will compete against