27th July 2026 – (Mountain View) Google has begun laying the groundwork for Gemini 4, signalling a step‑change in scale and ambition while its previously promised Gemini 3.5 Pro remains in partner trials. The company trailed the next generation alongside the rollout of Gemini 3.6 Flash, 3.5 Flash‑Lite and 3.5 Flash Cyber, framing Gemini 4 as its most demanding pre‑training run to date and a bid to stay competitive at the leading edge of artificial intelligence.
On Alphabet’s second‑quarter 2026 earnings call, chief executive Sundar Pichai said Google is both committed to and confident about operating at the AI frontier, but acknowledged gaps that still need closing, notably in coding and agentic coding. He argued that meaningful advances will require training substantially larger base models and described Gemini 4 as the vehicle for that leap rather than an incremental update. Pichai also indicated that Google intends to accelerate model releases to something approaching a monthly rhythm, with TPU capacity prioritised first for internal frontier‑level research before external customer demand.
The push comes as Gemini 3.5 Pro, introduced at Google I/O in May with a June target, continues partner testing without a firm launch date. Reporting has suggested the holdup stems from coding performance lagging internal expectations relative to rival systems, a dynamic Pichai tacitly addressed by emphasising renewed focus on developer‑centric capabilities and agentic workflows. Alphabet has lifted full‑year capital expenditure guidance to around $205 billion, underscoring sustained investment in the AI stack even as shares dipped after the results.
While Google highlighted lighter‑weight releases to show momentum—Gemini 3.6 Flash improved coding benchmarks while lowering token use, with 3.5 Flash‑Lite and 3.5 Flash Cyber aimed at high‑volume and security tasks respectively—the company’s narrative has shifted towards Gemini 4 as the next flagship. Based on previous cycles, industry watchers expect a late‑year debut, potentially in November or December, though Google has offered no timetable. The strategy, as outlined, is to establish a larger, frontier‑class baseline that can support faster, iterative updates across the portfolio.