Google defends Gemini 4 as employees report real-world struggles
Google unveiled its flagship Gemini 4 Argon model with a 1M-token output limit and pricing of $4 per 1M input tokens. Bloomberg reports that internally, employees say the model's real-world coding performance doesn't always match its benchmark scores, though Google denies underperformance.
- Gemini 4 Argon supports output of up to 1M tokens
- Pricing: $4 per 1M input and $20 per 1M output tokens
- Employees say the model struggles with certain coding tasks
- Google denies Gemini 4 underperforms in coding
Read next
AI