Google's highest-effort agentic research model, priced at $2.00/$12.00 per million tokens with a 1M-token context window and image output.
Recorded modality support shows what this model can accept and produce. An undocumented modality is shown as unknown rather than assumed to be unsupported.
Accepts
Produces
deep research max preview is Google's highest-effort agentic research model, built to run long, multi-step investigations and return a cited, synthesized report rather than a quick answer. It's the "max comprehensiveness" tier — the name signals that Google tuned it to spend more time and more tool calls per task than the standard research agent, at a price that reflects that: $2.00 per million input tokens and $12.00 per million output tokens, the highest rate of any model in this batch.
This isn't a chat model wearing a research label. It's built for tasks that involve breaking a question into sub-investigations, pulling from multiple sources, and assembling a coherent write-up with citations — work that takes many internal steps before it produces a final answer. Reasoning and function calling are both supported, which is what lets it plan and execute that kind of multi-step research loop rather than answer from a single pass.
A 1,048,576-token window matters more here than in a chat model because a research run can accumulate a large amount of intermediate context — search results, document excerpts, prior reasoning steps — before it produces final output. At $12.00 per million output tokens, a long, well-cited report with 5,000 output tokens costs around six cents on the output side alone, before counting the input tokens consumed by every intermediate step. Research-agent workloads accumulate cost across many internal calls, not just the one the user sees, so budgeting by cost per finished report rather than cost per token is the more honest way to plan around this model.
Input modalities are text, image, video, audio, and pdf; output is text and image — the image output is notable, since most models in this category answer in text only. That makes it a plausible fit for research tasks that should end in a chart, diagram, or annotated figure rather than a wall of prose.
Google also lists Gemini Deep Research Preview, which shares this model's context window, pricing, and modality support almost exactly. The "Max" name is the differentiator that matters here: treat Deep Research Max Preview as the variant to reach for when a task genuinely needs the deepest, most exhaustive pass Google offers, and the standard Deep Research Preview as the default for research tasks that don't need that extra depth.
The "Preview" label is worth taking seriously — this is not a model Google has committed to as a stable, long-term offering yet. Teams building production pipelines on it should expect the possibility of changes before it graduates to a general-availability name.
Use the recorded facts as a shortlist, then validate the model against representative inputs and production constraints.
Deep Research Max Preview is categorized for Any-to-any, Audio-text-to-text, Automatic speech recognition, Image-text-to-text, Image-to-image, Image-to-text, Text generation, Text-to-image, Text-to-text AI models, Video-text-to-text. Its current record accepts text, image, video, audio, pdf and produces text, image. Confirm file formats, preprocessing, and provider-specific request schemas before implementation.
The directory records 1.0M context and 65.5K maximum output. Listed token prices are $2.00 input and $12.00 output per 1M tokens. Treat missing values as unknown and recheck current commercial terms.
Recorded output speed is — and time to first token is —. Hosting route, region, prompt length, concurrency, and provider load can materially change both measurements.
This page separates sourced model facts from incomplete fields. Benchmark evidence is displayed only when its source is linked and verified. Before choosing Deep Research Max Preview, test task quality, tool reliability, safety behavior, data controls, rate limits, and total cost on the exact route you intend to use.
Related approved models in Text-to-image. Compare specifications and verify fit for your workload.
Google's agentic research model for multi-step investigation, priced at $2.00/$12.00 per million tokens with text and image output.
Image model for prompt-driven generation, editing, and visual design workflows
Image model for prompt-driven generation, editing, and visual design workflows
Fastest, most cost-efficient Gemini image model for high-volume 1K generation and editing
Nano Banana Pro for higher-fidelity image generation and design-heavy edits