Google DeepMind put out three models on July 21: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber. No update to Gemini 3.5 Pro, despite months of people asking for one. Google Cloud's Logan Kilpatrick said the team is testing 3.5 Pro with partners now and hopes to ship it soon, and that pretraining has already started on Gemini 4.
Of the three, 3.6 Flash is the one most developers will actually touch day to day. It's the mid-tier workhorse: cheaper and faster than a Pro-class model, but still built for agentic coding and long multimodal tasks.
Pricing and context window
Google's API pricing sets Gemini 3.6 Flash at $1.50 per million input tokens and $7.50 per million output tokens. The context window is 1,048,576 input tokens with up to 65,536 output tokens. It reads text, images, video, audio, and PDFs, though output is text only. The knowledge cutoff moved from January 2025 to March 2026.
| Gemini 3.6 Flash | Gemini 3.5 Flash-Lite | Gemini 3.5 Flash Cyber | |
|---|---|---|---|
| Input price | $1.50 / M tokens | Lower; billed as the cheapest in the lineup | Not public |
| Context window | 1.05M in / 65K out | Not disclosed | Not disclosed |
| Access | General API | General API | Governments and trusted partners only, limited pilot |
| Best for | Agentic coding, long multimodal jobs | High-volume, low-cost calls | Vulnerability research |
Benchmarks: efficient, not a leap
Google's own numbers show real gains over the previous Flash model: 49% versus 37% on DeepSWE, 83.0% versus 78.4% on OSWorld-Verified, 63.9% versus 49.7% on MLE-Bench, and a GDPval-AA v2 score of 1421 Elo versus 1349.
That's Google grading its own homework, though. Artificial Analysis, which runs independent evaluations, scores 3.6 Flash at 50 on its Intelligence Index: tied exactly with 3.5 Flash, not ahead of it. Community coding benchmarks put it behind GPT-5.6 Luna, Grok 4.5, and Kimi K3. The model does use roughly 17% fewer output tokens than its predecessor on the same Artificial Analysis testing, which is the actual headline here.
So this release is a cost story, not a capability jump. If you're already running high-volume agent workloads on 3.5 Flash, the token savings are real money at scale. If you were hoping for a model that finally beats [Grok 4.5](https://questloops.com/blog/grok-4-5-is-out-pricing-benchmarks-and-how-to-try-it-free) or closes the gap with [Claude Opus 5](https://questloops.com/blog/claude-opus-5-explained-pricing-benchmarks-and-where-it-actually-wins), this isn't that model.
What this means if you're already using Gemini
If you're on Gemini through the free tier or [using Gemini for free through the API](https://questloops.com/blog/how-to-use-google-gemini-for-free-in-2026-api-limits-explained), the upgrade to 3.6 Flash should show up automatically in most integrations without you doing anything. Google has also rolled Gemini into consumer products like [Gemini Notebook](https://questloops.com/blog/notebooklm-is-now-gemini-notebook-what-changed-and-what-didn-t) (the renamed NotebookLM), so the model swap quietly touches more than just API users.
Should you switch
Switch if you're on 3.5 Flash and care about cost per token; the drop in output tokens for equivalent quality adds up fast on agent workloads that run all day. Don't switch expecting a jump in raw intelligence: on independent benchmarks it's a lateral move, not an upgrade. And if your workload actually needs frontier-level reasoning, this generation of Flash still isn't the tool for that; you want Gemini 3.5 Pro once it ships, or a competitor's flagship model in the meantime.
FAQ
**Is Gemini 3.6 Flash free to use?** It's available through the Gemini API at the pricing above, and through consumer Gemini apps that include a free tier with rate limits.
**Is Gemini 3.6 Flash better than Gemini 3.5 Flash?** It's cheaper to run and faster on agentic and multimodal tasks by Google's own benchmarks, but independent testing shows it tied with 3.5 Flash on general intelligence, not ahead of it.
**When is Gemini 3.5 Pro coming?** Google says it's in partner testing with no confirmed ship date as of late July 2026.
