Skip to content
embodied.guide

Back to the AI models

Google DeepMindUnited Kingdom2026

Gemini 4 Argon

An AI model: software that a robot uses to perceive, plan or move. This page describes the software, not a robot.

RobotsSpotby Boston Dynamics B

Key figures as stated by the developer

4 of 6 stated · 2 gaps
  • Model type ⁠B general-purpose model
  • Controls not stated
  • Inputs ⁠A camera language audio
  • Access ⁠A API only
  • Parameters not stated
  • Base model ⁠A Not a fine-tune of a prior model
How to read the figures
50HzA
The letter in the grey square after a figure points to the source the figure came from. Pages with several sources carry several letters.

The four states

Versions and notes in the key figures (6)

Model type

Developer wording: "As a general-purpose physical reasoning model, Argon can be applied across the entire robotics lifecycle."

Controls

Note Model card, Outputs: "Text, with up to a 1M token input + output."

Inputs

Model card, Inputs: "Text strings (e.g., a question, a prompt, document(s) to be summarized), images, audio, and video files, with a combined input-output token context window of up to 1M." Images and video = camera, text = language.

Access

Model card, Distribution: "Our models are available to downstream providers via an application program interface (API) and subject to relevant terms of use."

Parameters

Note Model card, Architecture: "a large scale sparse mixture-of-experts (MoE)".

Base model

Model card, Model dependencies: Gemini 4 Argon, a new frontier model, "is not a modification or a fine-tune of a prior model".

Full specifications

6 of 12 fields stated
The letter points to the source at the foot of the page. Fields without a figure are dashed: the developer does not state them — that is not the same as zero, and not the same as no.
Field Value Source and caveats
Input and output
Model typegeneral-purpose modelB

Developer wording: "As a general-purpose physical reasoning model, Argon can be applied across the entire robotics lifecycle."

Inputscamera language audioA

Model card, Inputs: "Text strings (e.g., a question, a prompt, document(s) to be summarized), images, audio, and video files, with a combined input-output token context window of up to 1M." Images and video = camera, text = language.

Controlsnot stated

Note Model card, Outputs: "Text, with up to a 1M token input + output." ⁠A

Degrees of freedom controllednot stated

Note Robotics post, Spot demo: "Argon controls a Boston Dynamics Spot® quadruped (12-DoF legs + 6-DoF arm) to learn a new skill". ⁠B
Size and speed
Parametersnot stated

Note Model card, Architecture: "a large scale sparse mixture-of-experts (MoE)". ⁠A
Where it runs and what it is built on
Runs wherenot stated

Note Model card: "There is no required hardware or software to use the model." ⁠A

Base modelNot a fine-tune of a prior modelA

Model card, Model dependencies: Gemini 4 Argon, a new frontier model, "is not a modification or a fine-tune of a prior model".
Openness
AccessAPI onlyA

Model card, Distribution: "Our models are available to downstream providers via an application program interface (API) and subject to relevant terms of use."
The developer’s own claims
Learns new taskswithout examplesB

Developer claim: the robotics post lists "Mobile manipulation with MuJoCo practice, zero-shot robot control, and closed-loop steering of video world models". Its arm demo: "Argon controls a simple robotic arm to place a fallen mug upright and put the crumpled napkin inside."

Works across robotsyesB

Developer claim: the post shows control of two different robots: "Argon controls a Boston Dynamics Spot® quadruped (12-DoF legs + 6-DoF arm) to learn a new skill"; "Argon controls a simple robotic arm to place a fallen mug upright and put the crumpled napkin inside."

Yes only if the developer itself says the model runs on different robot models, for example robots from two manufacturers or two of one manufacturer’s models. Several units of the same robot do not count.

Not stated: Control rate, License

Notes

  • Google calls Gemini 4 Argon its frontier model for complex workflows in software engineering, enterprise knowledge work and cybersecurity defense; its robotics post presents it as a general-purpose physical reasoning model.
  • Google made Gemini 4 Argon available to developers on 13 October 2026.
  • Google DeepMind's registered office (Companies House, company 07386350) is in London.

Sources

One line per source, not per figure: a source has one retrieval date. The letter beside a figure points to the line it came from.