Gemini 4 Argon: Google makes its new frontier model globally available
Google DeepMind has released Gemini 4 Argon and made it globally available via API: introductory pricing of $2 per million input and $10 per million output tokens. What is documented, what remains a vendor claim — and how enterprises should check availability.
CogLake Editorial · 13 October 2026 · 4 min
Transparency: This article was automatically generated with the assistance of AI. All statements are based on the primary sources linked below (retrieved October 13, 2026); performance figures are vendor statements where noted.
Gemini 4 Argon: Google makes its new frontier model globally available
In one sentence: On October 13, 2026, Google DeepMind released the frontier model Gemini 4 Argon and made it available to developers and enterprises via API — at an introductory price of $2 per million input tokens and $10 per million output tokens.
What happened today
Google published two posts today: the model introduction (RSS timestamp: 16:01 UTC) and a robotics-focused companion post. We have not reviewed the latter in detail for this article.
Key points from the introduction:
- Availability: globally via API in AI Studio and on the Gemini platform; for Google AI Ultra subscribers in Google Antigravity; for AI Ultra and AI Pro subscribers in the Gemini app and AI Mode in Google Search.
- Positioning: per Google, built for “deep work” on complex, multi-step tasks across coding, finance, legal, cybersecurity, and science.
- Pricing: introductory $2 per million input tokens and $10 per million output tokens; cached input tokens are 95 percent cheaper. Google explicitly calls the price “introductory”; no end date is given.
Pricing and access channels (as of October 13, 2026)
| Channel | Access for | Pricing per Google |
|---|---|---|
| API: AI Studio, Gemini platform | Developers and enterprises, global | introductory $2/M input tokens, $10/M output tokens; cached input 95% off |
| Google Antigravity | Google AI Ultra | included in subscription |
| Gemini app and AI Mode in Search | Google AI Ultra and Pro | included in subscription |
The vendor-linked pricing page is authoritative; check current figures before budgeting.
What is documented — and what remains a vendor claim
| Statement | Source/status |
|---|---|
| Release and global API availability on October 13, 2026 | Primary source: Google blog (verified) |
| #1 on the Vals Index (finance, coding, legal, tax) | Vendor claim; index linked in the announcement |
| +35% agentic code submissions, roughly one third fewer rollbacks at Google teams | Vendor statement from internal use |
| Chromium refactoring: separating a ~6 million-line folder from a 40 million-line codebase | Vendor example (Chromium issue linked) |
| Deployments at Wiz (“Scan for Good”) and Fairwind program expansion | Vendor statements on customer use |
We did not independently verify the Vals ranking; benchmarks are a pointer, not a substitute for your own use case.
Four levels of model availability — status for Argon
| Level | Status (October 13, 2026) |
|---|---|
| 1. Vendor announcement | met — Google blog, October 13, 2026, 16:01 UTC |
| 2. Access at the provider | met — API globally, introductory pricing published |
| 3. Connection to your own platform | open — Google Gemini is among the supported providers; whether and when a specific model becomes selectable depends on provider availability and the release in your deployment |
| 4. Actual usability | check — model list, context window, tool support and pricing in your own account |
What enterprises can do now
- Budget with care: the introductory price makes pilot projects attractive; plan for price changes and calculate with the current provider pricing page.
- Verify availability instead of assuming it: an announcement is not proof of availability in your deployment. Check the model list and terms in your account or platform.
- Choose models per task: frontier models like Argon target multi-step deep work; for high-volume tasks smaller models may remain more economical — model choice is a per-task decision. CogLake supports Google Gemini alongside other providers and controlled BYOK (model overview).
Sources
- Google / Google DeepMind: “Gemini 4 Argon: Built to solve your hardest problems”, October 13, 2026 (RSS: 16:01 UTC), retrieved October 13, 2026 — https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/
- Google DeepMind RSS feed, retrieved October 13, 2026 — https://deepmind.google/blog/rss.xml
- Google AI Studio pricing page (linked in the announcement) — https://aistudio.google.com/docs/pricing
- Vals AI: Vals Index (linked in the announcement) — https://www.vals.ai/benchmarks/vals_index
- Google DeepMind: “Gemini 4 Argon for Robotics”, October 13, 2026 (title and date checked only) — https://deepmind.google/blog/gemini-4-argon-for-robotics/