Technology
Google unveils Gemini 4 Argon with expanded output capacity amid performance debate
The company introduced its latest flagship AI model with expanded limits as reports emerge of internal doubts over coding.
The short version
- Google announced Gemini 4 Argon, a new frontier AI model featuring an output capacity of up to 1 million tokens.[CNET · 9to5Google]
- Access is currently limited to trusted testers and cybersecurity specialists participating in Google's Fairwind program.[CNET · 9to5Google]
- A Bloomberg report noted that some internal staff found the model struggles with real-world coding, though Google denied underperformance claims.[9to5Google]
Key facts
- Google introduced Gemini 4 Argon as its newest frontier model on September 30, 2026.[CNET · 9to5Google]
- The model features a 1-million-token output capacity limit.[CNET · 9to5Google]
- Google stated the model achieved a 77.9% score on the DeepSWE v1.1 benchmark.[9to5Google]
- Access is presently limited to trusted testers and cybersecurity defenders in Google's Fairwind program.[CNET · 9to5Google]
- API pricing will debut at $2 per million input tokens and $10 per million output tokens before rising to $4 and $20, respectively.[9to5Google · 9to5Google]
What remains uncertain
- Whether Gemini 4 Argon struggles in real-world deployment, as reported by Bloomberg citing internal staff, or maintains frontier coding performance as Google claims.[9to5Google]
- The timeline for broader general availability beyond Fairwind program defenders and select testers.[CNET]
Sources
Outlet counts describe coverage, not independent confirmation. Reports may share a wire service or original source.