The “which launches first” question is settled: GPT-5.6 won, reaching full general availability on July 9, 2026. Gemini 3.5 Pro is now the one everyone’s waiting on, with July 17 circulating as the rumored launch date — though Google has not officially confirmed it. Here’s where things actually stand, updated with what’s confirmed versus still leaked.
Quick comparison: what’s confirmed vs rumoured
| Spec | Gemini 3.5 Pro | GPT-5.6 Sol |
|---|---|---|
| Developer | Google DeepMind | OpenAI |
| Status as of July 17, 2026 | Not yet released — July 17 rumored, unconfirmed by Google | Generally available since July 9, 2026 |
| Context window | 2 million tokens (leaked, unconfirmed) | ~1.05 million tokens (confirmed) |
| Max output | Not disclosed | 128,000 tokens (confirmed) |
| Reasoning mode | Deep Think, gated behind $250/mo Ultra plan (leaked) | Max and Ultra reasoning effort settings (confirmed) |
| Pricing | ~$1.25 input / $10 output per 1M tokens (leaked, unconfirmed) | $5 input / $30 output per 1M tokens (confirmed) |
| Last official signal | Sundar Pichai, Google I/O, May 19: in internal use, “next month” | OpenAI, July 9: full GA across ChatGPT, Codex, API |
What actually happened with GPT-5.6
GPT-5.6 launched as a government-gated limited preview on June 26, restricted to roughly 20 vetted organizations while the Commerce Department ran a 30-day cybersecurity review under the administration’s AI executive order. That review cleared early, and GPT-5.6 Sol, Terra, and Luna went fully public on July 9 — alongside ChatGPT Work, OpenAI’s new agentic productivity product. Confirmed pricing: Sol at $5/$30 per million tokens, Terra at $2.50/$15, Luna at $1/$6. Full GPT-5.6 GA breakdown here.
Where Gemini 3.5 Pro stands
Google scrapped the original Gemini 3.5 Pro base model and restarted pretraining, which is the reported reason for the slip from a June target into July. July 17 — the same day Shanghai’s World AI Conference opens with Xi Jinping attending in person — is the date circulating in leaks, with a 2-million-token context window, Deep Think reasoning gated behind the $250/month Ultra tier, and API pricing rumored near $1.25 input / $10 output per million tokens. None of this is confirmed by Google. Treat the date and every spec as unverified until an official model card lands. See our Gemini 3.5 Pro release tracker for the latest.
Frequently asked questions
Has Gemini 3.5 Pro launched yet?
Not as of this update. July 17, 2026 is the widely rumored date, coinciding with Shanghai’s World AI Conference, but Google has not published an official model card, pricing page, or confirmed launch date. Some reports flag July 24 as a fallback window if the date slips again.
Is GPT-5.6 available to everyone now?
Yes. As of July 9, 2026, all three tiers — Sol, Terra, and Luna — are generally available through ChatGPT, Codex, and the OpenAI API, with confirmed pricing and no access restrictions remaining.
Which one should I wait for?
You don’t have to wait for GPT-5.6 — it’s live now, and Terra offers GPT-5.5-class quality at roughly half the price for most production use. If your priority is a massive context window for large documents, codebases, or video, Gemini 3.5 Pro’s rumored 2M-token window is worth waiting on — but only once Google confirms it’s real. Neither replaces Claude Opus 4.8 for pure reasoning and writing quality; see our full three-way comparison with Claude Opus 4.8 for the complete picture.
What about Grok 4.5 and GLM-5.2?
Grok 4.5 launched shortly before GPT-5.6 and reset the value floor at $2 input / $6 output per million tokens. Zhipu’s GLM-5.2 is the open-weight option to watch, landing close to closed-source leaders on coding at a fraction of the cost and available to everyone right now — unlike Gemini 3.5 Pro.
This page is updated as Gemini 3.5 Pro’s launch develops. Last updated: July 17, 2026.
Related reading
- Gemini 3.5 Pro release date tracker — every confirmed spec and pricing detail
- GPT-5.6 GA launch breakdown — pricing, benchmarks, ChatGPT Work
- GPT-5.6 vs Claude Opus 4.8 vs Gemini 3.5 Pro — the full three-way model comparison