Gemini 3.6 Flash

What it is

Gemini 3.6 Flash is Google DeepMind’s workhorse model, launched on 21 July 2026. Flash is the tier built for high-volume everyday work rather than the hardest reasoning, and Google describes 3.6 as delivering “better coding, knowledge work, and multimodal performance” than the version before it. The pitch is efficiency: the same class of task done for less. Google says it “reduces output token usage by 17 per cent compared to 3.5 Flash,” and its price is lower too, at $1.50 per million input tokens and $7.50 per million output.

The numbers

A benchmark is a fixed set of tasks run to produce a comparable score, and the party publishing a launch benchmark has a stake in the outcome, so single figures are best weighed against independent testing. On Google’s own numbers, Gemini 3.6 Flash improves clearly on 3.5 Flash: 49 per cent against 37 on the DeepSWE coding benchmark, 63.9 against 49.7 on MLE-Bench, 83.0 against 78.4 on OSWorld-Verified for computer use, and 1421 against 1349 on the GDPval-AA measure. The pattern is a solid step up in coding and knowledge work while spending fewer tokens.

The tier that did not move

The more revealing detail is what Google did not ship. There is no Gemini 3.6 Pro, and Gemini 3.5 Pro has itself not yet reached general availability, with Google saying “Gemini 3.5 Pro is currently testing with partners and we plan to make it broadly available as soon as it’s ready.” So the 3.6 label applies to the workhorse tier, not the flagship. It is a reminder that a version number attaches to a specific model in a family, not to the whole line at once, and that reading “Gemini 3.6” as a single new flagship would be wrong.

Why it stands out

Flash is where most real traffic runs, because cost per task decides what is affordable to build. A workhorse model that does more while costing less pushes that line in the developer’s favour, which is why an efficiency-focused Flash release can be as consequential as a flashier flagship. What to watch is when the Pro tier catches up to the 3.6 generation, and how the Flash economics compare with rival workhorse models. For that comparison, see our large language models hub and the wider AI section.