Post-Cutoff

GoogleGemini familyPreviewLanguage modelReleased Apr 2026

Gemini Deep Research agent (API)

Google’s Deep Research agent exposed to developers through the Interactions API.

Model iddeep-research-preview-04-2026

Price$1 per task estimate low

Checked25 October 2026

No published training cutoff

At least1,780events since its release in April 2026

81 historic, 405 major. 193 days of news.

Google doesn’t publish a training cutoff, so we count from its release date. The real gap is larger.

Five things it has never heard of

How to call it

2 ways to use Gemini Deep Research agent (API), with the model id each provider expects.

Providers, model ids, endpoints and docs for Gemini Deep Research agent (API)
ProviderModel idEndpointDocs
Gemini API (Interactions API) deep-research-preview-04-2026 https://generativelanguage.googleapis.com/v1beta/interactions ai.google.dev
Gemini API (Interactions API), Max tier deep-research-max-preview-04-2026 https://generativelanguage.googleapis.com/v1beta/interactions ai.google.dev

Pricing

Prices for Gemini Deep Research agent (API)
Per task estimate low$1
Per task estimate high$3
Max per task estimate low$3
Max per task estimate high$7

As published: “USD per research task, Google’s own estimate (Deep Research ~80 searches, ~250k input / ~60k output tokens; Max up to ~160 searches, ~900k in / ~80k out). Billed pay-as-you-go at the underlying Gemini model’s rates plus tool fees; ‘estimates based on preview rates and subject to change’”.

Source: ai.google.dev, checked 25 October 2026.

Specs

Input
Text
Output
Text
Context window
Not published
Max output
Not published
Open weights
No
Licence
Proprietary
Training cutoff
Not published
Released
April 2026
Status
Preview
Type
Language model

What stands out

What its maker marketed at launch, and what people found later, each with its source.

  • Autonomous multi-step web research as an API agent

    Runs iterative searching and reading for several minutes and returns a synthesized report; only via the Interactions API with background=true (poll or stream), not generate_content. ‘Max’ trades speed for comprehensiveness.

    Source: ai.google.dev

Notes

Agent ids, not raw models; the underlying Gemini model is not named on the docs page; it gives per-task cost estimates instead (checked 2026-10-25). The older deep-research-pro-preview-12-2025 was deprecated on 2026-10-08 and shuts down on 2026-10-23 (Gemini API changelog); migrate to deep-research-preview-04-2026 or deep-research-max-preview-04-2026.

See the docs page for request examples (background execution is required).

Sources: https://ai.google.dev/gemini-api/docs/deep-research , https://ai.google.dev/gemini-api/docs/changelog

Not yet checked against an official source: The ‘04-2026’ suffix suggests an April 2026 snapshot; exact release day not verified.