Gemini 4 Argon hits general availability, free tiers left out
Google spent months as the chase car in the frontier race — today it put its flagship model in buyers' hands, while one of Anthropic's founders moved up his own timeline for when the field starts improving itself.
Google released Gemini 4 Argon to the public today — its first frontier model since Gemini 3, and the one it needs to stop trailing Anthropic and OpenAI. The rollout reaches developers through the AI Studio and Google Cloud APIs, Antigravity for Google AI Ultra subscribers, and the Gemini app and Search's AI Mode for Pro and Ultra subscribers; The Verge reports it is not clear when — or if — Plus and free tiers will ever get it. Google claims Argon beats GPT-6 Astra, Claude Fable 5.1, and Claude Opus 5.5 across coding, cybersecurity, science, finance, and law benchmarks and ranks first on the Vals Index, and it says internal users have already lifted agentic code submissions more than 35% in Antigravity while cutting rollbacks by nearly a third, plus roughly halving the compute running YouTube's comments section. The company also published today's robotics push: Argon scores 87% on the ERQA embodied-reasoning benchmark, and in an internal test it learned a physical skill from a raw video of a person playing cornhole well enough to sink seven consecutive shots on Apptronik's Apollo humanoid.
The specs were part of September's tease — Google's Gemini 4 Argon gets a 1M-token output limit — but the real news today is availability, not paper, and the interesting tension is pricing versus access: at the introductory $2 per million input tokens and $10 per million output, The Verge puts Argon at roughly a fifth of the cost of GPT-6 Astra and Fable 5, yet the cheapest way to actually touch Google's best model is a paid subscription. Landing a day after Meta's Muse Wonder Full, the launch is a reminder that the frontier is still shifting under everyone's feet.
Anthropic co-founder Jack Clark now thinks full recursive self-improvement is more likely to become possible in 2027 — a year earlier than he previously thought. In the latest Import AI, Clark wrote that "I now think it's more likely we'll see full RSI become possible in 2027, whereas before I thought it was more likely in 2028," pointing at new benchmark evidence that frontier models' research taste — the ability to judge and conduct AI research — has been doubling roughly every three months since December 2025, with Opus 5.5 beating a 24-expert human baseline drawn from OpenAI, DeepMind, Nvidia, and Microsoft at about one-thirtieth the compute cost. It is a striking acceleration next to the caution Google's own chief scientist recently urged — Google won't confirm the RSI rumor. Its chief scientist just backed a slowdown.. Commenters pushing back on the thread note the passive framing is doing heavy lifting: if RSI is arriving, someone is actively building it.
Workday is making Sana the new front door of its platform — the startup it bought last year becomes the default interface for AI across HR and finance. Announced at Workday Rising, the "Workday Home" is available now, with the Sana Concierge assistant for HR and IT tasks expected in the first quarter of 2027, across a customer base of more than 11,500 organizations including over 65% of the Fortune 500. Workday completed its roughly $1.1 billion acquisition of Sana last November, so today's news is the integration rather than the deal — though Workday did add a fresh one, agreeing to buy the agentic-identity startup ConductorOne the same day. The philosophy, in Sana's Joel Hellermark's words: "AI is the new UI" — users should not have to decide which agent to use.
What to watch: whether Plus and free tiers ever get Argon, and whether independent evaluations back Google's claims over Astra and Fable.
Is a flagship model you can only try behind a paid subscription still the frontier — or just a press release? Tell us in the comments.
