Announcing Gemini 4 Argon

Google announced Gemini 4 Argon on October 13, 2026 — its latest frontier model, built for complex, multi-step work in coding, finance, legal, cybersecurity, and science. It combines reasoning, visual perception, domain knowledge, and stronger writing. Gemini 4 Argon is already available on vega.chat in the Ultimate model category, with vision and PDF inputs, and through api.vega.chat.

What Gemini 4 Argon does

Google's announcement focuses on work that takes more than a single answer: navigating large codebases, synthesizing technical documents, carrying out extended research. What stands out:

Multi-step reasoning across professional domains. Google reports a #1 ranking on the Vals Index, which covers finance, coding, legal, and tax work. Through the Gemini API, Argon runs with a unified 1 million-token input-output limit via a new continuation protocol.

Agentic software engineering. Google's own teams increased agentic code submissions by more than 35% while cutting code rollbacks by nearly one-third. In Chromium, Argon helped separate core webpage handling from a roughly 6-million-line legacy folder, compressing months of work into weeks.

Science and optimization. In the xTB quantum chemistry package, Google reports up to 2.3× faster molecular-structure prediction and 1.8× faster conformer search. Argon also produced a reported 2.85× end-to-end speedup for Gemma 4 E2B inference on an A100 with vLLM.

Document synthesis. Isomorphic Labs reportedly reduced PDF and patent synthesis from weeks to hours.

Robotics. Argon scores 87% on ERQA. In an internal demo it turned a cornhole video into simulation code and robot movements — Apptronik's Apollo humanoid then made seven consecutive shots.

Writing. Google describes better adherence to style instructions, more direct answers, and less repetition than previous Gemini models.

These are Google's reported results; real numbers will vary by task.

Availability and pricing

Developer access goes through the Gemini API paid tier in AI Studio and Android Studio; Google AI Ultra subscribers also get Antigravity. Enterprise channels are Gemini for business, the API, and Antigravity. Consumers get Argon in the Gemini app and AI Mode in Search with a Google AI Ultra or Pro subscription.

Google's introductory Gemini API pricing is $2 per million input tokens and $10 per million output tokens, with a 95% discount on cached input. The introductory period ends November 30, 2026 — after that, the listed rates become $4 per million input and $20 per million output. These are Google's API rates, not VEGA pricing.

Google is also opening a research program for 10,000 scientists — Antigravity's highest usage tier free for six months — and expanding its Fairwind Program for cybersecurity defenders to hundreds of new partners.

Try Gemini 4 Argon on vega.chat

Sign in to vega.chat and pick Gemini 4 Argon in the Ultimate category. Vision and PDF inputs are supported: try a document-synthesis task, ask questions about an image, or hand it a detailed writing brief. Developers can also reach Argon through the api.vega.chat OpenRouter-compatible API — for agents, coding workflows, and their own applications.

Announcing Gemini 4 Argon

What Gemini 4 Argon does

Google's announcement focuses on work that takes more than a single answer: navigating large codebases, synthesizing technical documents, carrying out extended research. What stands out:

Multi-step reasoning across professional domains. Google reports a #1 ranking on the Vals Index, which covers finance, coding, legal, and tax work. Through the Gemini API, Argon runs with a unified 1 million-token input-output limit via a new continuation protocol.

Agentic software engineering. Google's own teams increased agentic code submissions by more than 35% while cutting code rollbacks by nearly one-third. In Chromium, Argon helped separate core webpage handling from a roughly 6-million-line legacy folder, compressing months of work into weeks.

Science and optimization. In the xTB quantum chemistry package, Google reports up to 2.3× faster molecular-structure prediction and 1.8× faster conformer search. Argon also produced a reported 2.85× end-to-end speedup for Gemma 4 E2B inference on an A100 with vLLM.

Document synthesis. Isomorphic Labs reportedly reduced PDF and patent synthesis from weeks to hours.

Robotics. Argon scores 87% on ERQA. In an internal demo it turned a cornhole video into simulation code and robot movements — Apptronik's Apollo humanoid then made seven consecutive shots.

Writing. Google describes better adherence to style instructions, more direct answers, and less repetition than previous Gemini models.

These are Google's reported results; real numbers will vary by task.