Overview
Gemini 3.6 Flash is Google's generally available workhorse model for agentic coding, knowledge work, and multimodal reasoning, released on July 21, 2026 with the API identifier gemini-3.6-flash. Google positions it as a production-ready upgrade over Gemini 3.5 Flash and Gemini 3 Flash Preview: it keeps the 1,048,576-token input window and 65,536-token output limit while reducing output pricing and improving reliability in multi-step agent loops.
The model supports text, image, video, audio, and PDF input with text output. It also supports caching, code execution, file search, function calling, Google Maps grounding, search grounding, structured outputs, thinking, URL context, and Computer Use in preview. It does not support image generation, audio generation, or the Live API.
What's New
Stronger agentic coding and execution
Gemini 3.6 Flash is tuned for longer coding and agent workflows where a model must inspect context, make changes, run checks, and avoid unnecessary edits. Google's latest-model guidance says 3.6 Flash reduces reasoning loops, unwanted file changes, tool calls, and conversational turns compared with Gemini 3.5 Flash. That makes it a better default for production coding agents, repository maintenance, and multi-step tool workflows where reliability matters as much as raw benchmark score.
Better multimodal and spatial reasoning
Google highlights improvements in chart interpretation, visual blueprint conversion, and multi-element web layout generation. The official model page lists support for text, image, video, audio, and PDF inputs, which keeps 3.6 Flash in the same broad multimodal family as 3.5 Flash while improving performance on visual and spatial tasks.
Lower output pricing than Gemini 3.5 Flash
Standard API pricing remains $1.50 per 1M input tokens, but output pricing drops from $9.00 per 1M output tokens on Gemini 3.5 Flash to $7.50 per 1M output tokens on Gemini 3.6 Flash. For generation-heavy agent workflows, that price change matters because the model is also designed to complete work with fewer turns and fewer tool calls.
Computer Use and built-in tools
Gemini 3.6 Flash supports Computer Use as a preview capability, alongside code execution, file search, function calling, search grounding, structured outputs, URL context, and caching. This makes it a stronger candidate for agents that need to inspect webpages, use applications, call tools, and produce structured results in one workflow.
Performance Benchmarks
Google's July 21, 2026 announcement and Gemini model pages report broad gains over Gemini 3.5 Flash and earlier Gemini 3 models. Key figures include:
| Benchmark | Gemini 3.6 Flash | Gemini 3.5 Flash |
|---|---|---|
| DeepSWE v1.1 | 49% | 37% |
| MLE-Bench | 63.9% | 49.7% |
| OSWorld-Verified | 83.0% | 78.4% |
| GDPval-AA v2 | 1421 | 1349 |
Google also says Gemini 3.6 Flash uses 17% fewer output tokens than Gemini 3.5 Flash on the Artificial Analysis Index. Treat these as Google-published launch benchmarks; teams should still validate quality, cost, and latency on their own prompts before routing production traffic.
Migration Guide
From Gemini 3.5 Flash
For most agentic coding, multimodal reasoning, and knowledge-work tasks, Gemini 3.6 Flash is the direct upgrade target. Update the model ID from gemini-3.5-flash to gemini-3.6-flash, rerun your regression prompts, and compare total task cost rather than only token price. The output-token rate is lower, and Google expects fewer turns and tool calls, but real savings depend on your prompts, tool schema, and retry behavior.
From Gemini 3 Flash Preview
Google lists Gemini 3.6 Flash as the recommended replacement for gemini-3-flash-preview on the deprecations page. Moving from preview to stable should reduce lifecycle risk, but you should still check API behavior changes, default thinking settings, and structured output assumptions before production rollout.
API changes to check
Google's latest-model guide notes that Gemini 3.6 Flash and Gemini 3.5 Flash-Lite deprecate sampling parameters such as temperature, top_p, and top_k, and no longer support prefilled model turns. Remove those fields from requests and test any conversation-state code that relies on a final model turn.
Compatibility Notes
Supported inputs and outputs:
| Type | Support |
|---|---|
| Text input | Supported |
| Image input | Supported |
| Video input | Supported |
| Audio input | Supported |
| PDF input | Supported |
| Text output | Supported |
| Image output | Not supported |
| Audio output | Not supported |
| Live API | Not supported |
Core model limits and features:
| Capability | Gemini 3.6 Flash |
|---|---|
| Model ID | gemini-3.6-flash |
| Lifecycle | Stable / GA |
| Input token limit | 1,048,576 |
| Output token limit | 65,536 |
| Default thinking level | medium |
| Computer Use | Supported (Preview) |
| Batch API | Supported |
| Flex inference | Supported |
| Priority inference | Supported |
Pricing & Plans
Gemini 3.6 Flash is available through the Gemini API with a free tier for eligible development usage and paid token pricing for production workloads.
| Tier | Input | Output | Notes |
|---|---|---|---|
| Free | $0 | $0 | Rate-limited development usage where available |
| Standard | $1.50 / 1M tokens | $7.50 / 1M tokens | Main published rate in Google's latest-model guide |
The model page also lists Batch API, Flex inference, and Priority inference as supported consumption options. Confirm current production rates on Google's Gemini API pricing page before committing large workloads, because capacity options and regional enterprise pricing can vary.
Best For
- Coding agents that need to inspect repositories, edit files, run checks, and avoid unnecessary changes.
- Multimodal workflows involving charts, screenshots, blueprints, documents, or visual layout reasoning.
- Production agents where total task cost depends on turns, retries, and tool calls, not just per-token pricing.
- Teams migrating away from Gemini 3 Flash Preview to a stable replacement.
- Applications that need built-in tools such as code execution, function calling, search grounding, URL context, or Computer Use Preview.
FAQ
What is the model ID for Gemini 3.6 Flash?
Use gemini-3.6-flash. Google's model page lists it as a stable model with a July 2026 latest update.
Is Gemini 3.6 Flash generally available?
Yes. Google's latest-model guide lists Gemini 3.6 Flash as generally available and ready for production use.
How much does Gemini 3.6 Flash cost?
Google lists Standard pricing at $1.50 per 1M input tokens and $7.50 per 1M output tokens. Free-tier availability, Batch, Flex, Priority, and enterprise pricing should be checked on the official Gemini API pricing page before production rollout.
Does Gemini 3.6 Flash support Computer Use?
Yes. The official model page lists Computer Use as supported in preview. It also supports code execution, file search, function calling, grounding, structured outputs, thinking, URL context, and caching.
Should I use Gemini 3.6 Flash or Gemini 3.5 Flash-Lite?
Use Gemini 3.6 Flash for agentic coding, complex multimodal reasoning, and higher-quality multi-step workflows. Use Gemini 3.5 Flash-Lite for high-throughput extraction, document processing, and latency-sensitive tasks where the lower $0.30 input price matters more than peak capability.
