gemini 3.6 flash: $1.50 per million tokens, 1m context, and three new models at once

google shipped three Gemini models at once on july 21, 2026: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber. the flagship of the batch is 3.6 Flash, which also landed in GitHub Copilot on the same day — making it one of the faster model-to-tool rollouts google has run.

what changed in 3.6 flash

the headline numbers: 1M-token context window, 64K output cap, and API pricing at $1.50 per million input tokens and $7.50 per million output tokens. the output price is lower than the previous 3.5 Flash rate — google says it's about 17% cheaper overall.

google's documentation notes “higher precision with fewer unwanted code edits and reduced execution loops” compared to 3.5 Flash. that kind of claim is hard to verify without a benchmark, and google hasn't published one that directly isolates that improvement. the model also ships with an updated knowledge cutoff: March 2026, which matters for questions about recent libraries, APIs, and tooling.

the other two models

3.5 Flash-Lite is positioned as the cheaper, faster option for tasks where latency and cost matter more than depth: classification, summarization, formatting. the pricing is below 3.6 Flash; exact numbers are in the Google AI Dev Docs.

3.5 Flash Cyber is security-focused — built and tuned for security analysis, vulnerability scanning, and threat assessment workloads. google is pitching it as a drop-in for teams already using Gemini for code review who want a model specifically optimized for security rather than a general-purpose one that also handles it.

what it means for api users

for teams building on the Gemini API, 3.6 Flash is now the default recommendation for coding-adjacent work. the pricing is lower than 3.5 Flash was at equivalent capabilities, the context window matches the frontier, and it's available immediately in Google AI Studio, Vertex AI, and the Gemini API.

the 17% cost reduction also compounds on volume. if your team runs high-throughput code generation, review, or analysis pipelines through the API — and was using 3.5 Flash — the switch to 3.6 Flash is a straightforward cost reduction. the only reason to stay on 3.5 Flash is if you've validated that its output format matches a specific downstream parser you've already built.

copilot angle

3.6 Flash is now selectable in Copilot for Pro, Pro+, Max, Business, and Enterprise plans. we covered the copilot-side of this — the credit meter and model picker context — in a separate piece: copilot's model picker now comes with a credit meter. the short version: if you're using Copilot, it's in your model picker now. if you're using the Gemini API directly, the upgrade path is a model string swap.

honest take

three models at once is a pattern google has used before — bundle the premium release with two cheaper variants so the announcement looks larger. the actual story is simpler: 3.6 Flash is a real upgrade to 3.5 Flash, it's cheaper, and the 1M context is now table stakes across the tier rather than a premium feature. if you're evaluating whether to build on gemini vs. alternatives at this price point, the cost-per-context math shifted in google's favor this week.