Gemini 4 Argon is Google's new flagship frontier model, released globally on October 13, 2026 for developers and enterprises through the Gemini API, for Google AI Ultra subscribers inside Google Antigravity, and for Ultra and Pro subscribers in the Gemini app and AI Mode in Search. It targets deep, multi-step work in domains that have been slow to adopt general-purpose models, with a stated focus on coding, finance, legal, cybersecurity, and science, where Google says it ranks first on the Vals Index that measures performance across those economic sectors.
The launch matters because it reframes how Google is positioning Gemini against OpenAI and Anthropic, pitching Argon less as a chat assistant and more as an engine for long-horizon engineering and research tasks. Google reports that teams using Argon in Antigravity have increased agentic code submissions by more than 35% while reducing code rollbacks by nearly a third , with internal case studies spanning a Chromium refactor across a roughly six million line legacy folder, overnight optimizations to the open-source Agent Development Kit, and speedups to the xTB quantum chemistry package. Pricing starts at $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted by 95%.
The release was announced by Sundar Pichai, with the model introduction authored by Koray Kavukcuoglu, SVP at Google DeepMind and Chief AI Architect, signaling that Argon is the production face of DeepMind's current frontier research rather than a standalone product line.
Comments (8)
Sign in to join the discussion.
N
Nneka O.19h ago
Argon? we went from Pro and Ultra to naming models after noble gases now. by Gemini 7 we'll be at Krypton and marketing will insist it's not a Superman reference.
M
maya r.19h ago
we had a model codenamed Argon internally in 2021 and shipped it to exactly zero users. glad to see the name finally getting its moment.
M
Milos Ravnic19h ago
the tweet buries the lede. 'available in paid tier, Ultra, Antigravity, Pro, app, Search' reads like a tier matrix someone lost a bet writing.
H
Habib K.19h ago
ok wait, Antigravity is the IDE and Argon is the model and Ultra is the tier. the naming committee is cooking something and I'm not sure it's dinner.
R
Rin Takahashi19h ago
fine but which SKU gets me a BAA, data residency in Frankfurt, and a procurement contact who answers emails before Q2. asking for 400 lawyers.
P
Priya Varghese19h ago
docs page loaded before the launch video finished buffering, which is honestly the most impressive benchmark here.
S
Søren Dahl19h ago
every one of these launches has a cherry picked coding demo and a vibes benchmark. wake me when I can run my actual eval suite without it hallucinating a function that doesn't exist.
L
Lucia Marchetti19h ago
'Built to solve your hardest problems' is what every model since GPT-2 has said. try 'Gemini 4 Argon. the one your intern won't break.'
Work With Us
Book an intro call to explore our suite of services. We might have a waitlist.
Google
Built to solve your hardest problems
Built to solve your hardest problems
Tags
TLVC Rating
Community Rating
No ratings yet
to rate this launch.
Gemini 4 Argon is Google's new flagship frontier model, released globally on October 13, 2026 for developers and enterprises through the Gemini API, for Google AI Ultra subscribers inside Google Antigravity, and for Ultra and Pro subscribers in the Gemini app and AI Mode in Search. It targets deep, multi-step work in domains that have been slow to adopt general-purpose models, with a stated focus on coding, finance, legal, cybersecurity, and science, where Google says it ranks first on the Vals Index that measures performance across those economic sectors. The launch matters because it reframes how Google is positioning Gemini against OpenAI and Anthropic, pitching Argon less as a chat assistant and more as an engine for long-horizon engineering and research tasks. Google reports that teams using Argon in Antigravity have increased agentic code submissions by more than 35% while reducing code rollbacks by nearly a third , with internal case studies spanning a Chromium refactor across a roughly six million line legacy folder, overnight optimizations to the open-source Agent Development Kit, and speedups to the xTB quantum chemistry package. Pricing starts at $2 per million input tokens and $10 per million output tokens, with cached input tokens discounted by 95%. The release was announced by Sundar Pichai, with the model introduction authored by Koray Kavukcuoglu, SVP at Google DeepMind and Chief AI Architect, signaling that Argon is the production face of DeepMind's current frontier research rather than a standalone product line.
Comments (8)
Argon? we went from Pro and Ultra to naming models after noble gases now. by Gemini 7 we'll be at Krypton and marketing will insist it's not a Superman reference.
we had a model codenamed Argon internally in 2021 and shipped it to exactly zero users. glad to see the name finally getting its moment.
the tweet buries the lede. 'available in paid tier, Ultra, Antigravity, Pro, app, Search' reads like a tier matrix someone lost a bet writing.
ok wait, Antigravity is the IDE and Argon is the model and Ultra is the tier. the naming committee is cooking something and I'm not sure it's dinner.
fine but which SKU gets me a BAA, data residency in Frankfurt, and a procurement contact who answers emails before Q2. asking for 400 lawyers.
docs page loaded before the launch video finished buffering, which is honestly the most impressive benchmark here.
every one of these launches has a cherry picked coding demo and a vibes benchmark. wake me when I can run my actual eval suite without it hallucinating a function that doesn't exist.
'Built to solve your hardest problems' is what every model since GPT-2 has said. try 'Gemini 4 Argon. the one your intern won't break.'