LLMs
Gemini 4 Argon Now Available: Price, Access, What Changed
Oct 14, 2026·5 min read·By Rahish Saifi
On October 13, 2026, Google ended the wait that began with its October 1 phased announcement and made Gemini 4 Argon globally available to developers, enterprises, and paying subscribers. In a post attributed to Koray Kavukcuoglu, SVP of Google DeepMind and Chief AI Architect at Google, the company framed Argon as a frontier model built for deep, multi-step work in coding, finance, legal, cybersecurity, and science—and said it ranks #1 on the Vals Index, which tracks performance across economically critical sectors.
If you followed our earlier coverage of the phased Argon rollout, broad API and consumer access was still labeled “coming soon.” That line is now concrete: Argon is rolling out worldwide through specific Google surfaces, with published API pricing and a hard deadline after which token costs double.
What changed since the October 1 announcement?
Google’s first Argon reveal emphasized a staged path: an initial rollout to trusted cyber defenders through the Fairwind Program, feedback from early testers, and safeguard testing before general availability. The October 13 launch is the flip to global GA for the paid developer API and major product integrations—not a new model, but a widening of who can use it.
Oct 1 had already announced a 1M output-token limit; the Oct 13 post adds the unified input–output framing, the continuation protocol, and specific numbers on agentic coding, open-source speedups and safety.
Who can use Gemini 4 Argon today?
Access paths differ by audience; Google’s posts are explicit about tiers:
- Developers and enterprises: Build via the paid tier of the Gemini API in AI Studio, the Gemini platform, and Android Studio. Enterprises can also use Gemini for business, the API, and Google Antigravity.
- Google AI Ultra subscribers: Argon in Google Antigravity and in AI Studio (per the developer post). Ultra is the Antigravity tier called out for agent-first development.
- Google AI Ultra and Pro subscribers: Argon in the Gemini app and AI Mode in Search—not limited to Ultra alone for everyday chat and search experiences.
- Cybersecurity defenders: Google is beginning expansion to hundreds of additional Fairwind Program partners; defenders can apply to join for access to Argon without cyber guardrails on a rolling basis.
- Scientists and researchers: Google is offering 10,000 scientists the highest Antigravity usage tier free for six months, onboarded on a rolling basis and subject to eligibility, capacity, and selection; applications are open.
There is no single “free for all” API tier in the launch posts—general developer access is tied to the paid Gemini API, while consumer access flows through Google’s subscription products.
Pricing: intro rates until November 30, 2026
For teams budgeting inference, the headline number is time-limited. Google lists an introductory price of $2 per 1M input tokens and $10 per 1M output tokens, with cached input tokens at 95% off. After the introductory period expires on November 30, 2026, the public Argon list price moves to $4 per 1M input and $20 per 1M output (Google notes this applies to the Argon model intended for public access).
If you are prototyping large-context pipelines or agent loops, the combination of the 1M token window and discounted cache reads is materially cheaper right now—but production cost models should assume the post–Nov 30 doubling unless Google extends the promo.
What Google says Argon is good at
Google positions Argon for “hard problems” that span repositories, documents, and tools—not quick chat alone. In its engineering examples, Argon in Antigravity helped raise agentic code submissions by more than 35% inside Google while reducing code rollbacks by nearly a third. Public anecdotes include Chrome’s multi-million-line refactor work, overnight performance work on the open-source Agent Development Kit (ADK) (Google cites 40% faster in-memory session prep in long sessions and 21% faster cold imports), and a 2.85× end-to-end inference speedup for Gemma 4 E2B on an A100 via vLLM with full token parity.
Google also cites xTB speedups (up to 2.3× in one workflow), an 87% ERQA embodied-reasoning score, and an internal Apptronik Apollo cornhole demo with seven consecutive shots after video-to-simulation coding.
Safety and agent behavior
Because Argon is aimed at agents and high-capability coding, Google’s October 13 post spends meaningful space on safeguards: shortcut attempts on coding benchmarks fell to 2% on Gemini 4 Argon; Gray Swan indirect prompt injection attack success across 15 adaptive attempts was cut to 0.7%; and external bio red teaming unlocked less than 1% of bio sub-tasks Google tests for. Misuse resistance and monitoring layers are part of the pitch—especially as Fairwind expands to more defenders.
What developers and users should do now
Developers on the API: Start in AI Studio or your existing Gemini API project on the paid tier, read Google’s developer guide for the continuation protocol, and model costs with both the intro and post–Nov 30 price sheets. If you build agents, pair Argon with your orchestration stack and test cache-friendly prompt patterns before the 95% cache discount matters less at higher list prices.
Ultra subscribers: Treat Google Antigravity as the primary surface for long-horizon coding and multi-agent loops; Google’s developer post points to Antigravity and Android Studio blogs for setup.
Pro or Ultra Gemini app users: Switch to Argon inside the Gemini app or AI Mode for complex document and analysis tasks Google showcases (legal, finance, multimodal decks)—you do not need Ultra for app access, but you do need a paid Google AI subscription tier.
Enterprises: Evaluate through Gemini for business and benchmark on your own workflows. For context on Google’s enterprise agent stack, see our Gemini at Work 2026 coverage or building your first AI agent in 2026.
Bottom line
Gemini 4 Argon’s October 13 launch turns Google’s frontier model from a phased, defender-first story into something you can actually bill and ship against—via paid API access, subscription surfaces, and enterprise channels. The near-term pricing is aggressive; the calendar is not. Mark November 30, 2026 for cost replanning, confirm which subscription tier unlocks Antigravity versus the Gemini app, and read Google’s own setup docs before relying on million-token runs in production.