Google releases Gemini 3.6 Flash and 3.5 Flash-Lite, with Flash Cyber in limited pilot

Google announced Gemini 3.6 Flash, Gemini 3.5 Flash-Lite and Gemini 3.5 Flash Cyber on July 21, 2026. The first two models are available today across several Google products and developer surfaces. Flash Cyber is not generally available: it is deployed only through a limited-access CodeMender pilot for governments and trusted partners.

What the three models actually do

Gemini 3.6 Flash is the workhorse of the release. Google positions it for coding, knowledge work and multimodal tasks. It costs $1.50 per 1 million input tokens and $7.50 per 1 million output tokens. That price sits above the new Flash-Lite model, so the practical question is whether its stronger benchmark results and lower token use justify the difference for a given workload.

Google reports that Gemini 3.6 Flash reduces output token usage by 17% compared with 3.5 Flash on the Artificial Analysis Index. In some DeepSWE cases, the reduction reaches 65%. Token reduction matters separately from the listed per-token price because a model that produces fewer output tokens may reduce the total cost of completing a task.

The reported benchmark changes compare the new model with the previous one:

• DeepSWE: 49% versus 37%.
• MLE Bench: 63.9% versus 49.7%.
• OSWorld-Verified: 83.0% versus 78.4%.
• GDPval-AA v2: 1421 versus 1349.

Gemini 3.5 Flash-Lite targets fast, inexpensive, high-throughput work. It costs $0.30 per 1 million input tokens and $2.50 per 1 million output tokens. Google cites 350 output tokens per second on the Artificial Analysis Index. Its reported benchmark comparisons are also higher than the previous model:

• Terminal-Bench 2.1: 54% versus 31%.
• GDM-MRCR v2: 72.2% versus 60.1%.
• GDPval-AA v2: 1140 versus 642.
• SWE-Bench Pro: 54.2% versus 49.6%.
• OSWorld-Verified: 74.0% versus 65.1%.

Gemini 3.5 Flash Cyber has a narrower purpose. It is a cybersecurity model for vulnerability detection and patching. Its restricted deployment reflects the model's dual-use nature. Google is making it available only through CodeMender, not through the general Gemini surfaces used by the other two models. Full details are in the Google announcement.

What this means when choosing tools

For teams already using Gemini, the release creates a clearer split between a general workhorse and a throughput-oriented option. Gemini 3.6 Flash fits coding, multimodal and knowledge tasks where the stronger reported results or reduced output length matter. Gemini 3.5 Flash-Lite fits workloads where input and output price, speed and volume are the primary constraints. The two price schedules make it possible to estimate the model-cost difference before testing quality on a representative task set.

Developers can access both general models through the Gemini API. Google AI Studio is one of the listed Gemini 3.6 Flash access points, alongside Android Studio. Teams comparing coding workflows can also place these results in the context of Gemini CLI, GitHub Copilot and Cursor. Those tools are not substitutes for benchmark evaluation: the relevant choice depends on the model access, interface and workflow a team needs. More options are collected in the Coding category.

Availability differs by product. Gemini 3.6 Flash is available in the Gemini API through Google AI Studio and Android Studio, the Gemini Enterprise Agent Platform, the Gemini app and Google Antigravity. Gemini 3.5 Flash-Lite is available in the Gemini API, Gemini Enterprise Agent Platform, the Gemini app and Google Search. This means both can be tested through the API today, while their additional distribution surfaces are not identical.

Limits and what is not available yet

Gemini 3.5 Flash Cyber is not an open API or broadly released product. Access is limited to the CodeMender pilot for governments and trusted partners. Buyers should not plan a general production integration around it unless they are part of that restricted program. The announcement establishes its vulnerability detection and patching focus, but does not make it another selectable Flash tier for ordinary Gemini API users.

Two future items are also announced but not shipped generally. Gemini 3.5 Pro is in partner testing, with general availability planned. Gemini 4 is earlier in the cycle: pre-training is underway. Neither statement amounts to availability today, and Google gives no generally available release in this announcement.

The immediate decision is therefore between Gemini 3.6 Flash and Gemini 3.5 Flash-Lite, both of which are live across the listed surfaces. The next points to watch are general availability for Gemini 3.5 Pro and whether Flash Cyber expands beyond CodeMender's limited-access deployment. Gemini 4 remains a development signal rather than a product teams can select today.