Gemini 4 Argon Expands Output to 1 Million Tokens

Gemini 4 Argon has a 1 million-token output limit, expanding the maximum output from the 64,000-token limit available in prior models. [1] Initially, Gemini 4 Argon costs $2 per 1 million input tokens and $10 per 1 million output tokens, while the later rates are set to rise to $4 per 1 million input tokens and $20 per 1 million output tokens. [1]

Artificial Analysis says Gemini 4 Argon in its high configuration matches GPT-6 Astra in its max configuration on the evaluator’s Intelligence Index. [2] In the same Artificial Analysis comparison, Gemini 4 Argon is reported with a 15% hallucination rate, whereas GPT-6 Astra is reported with a 51% hallucination rate. [2] The reported Intelligence Index result is a match between the named configurations, but the reported hallucination rates are 15% for Argon and 51% for Astra. [2]

Argon’s million-token output ceiling is the clearest practical departure from the 64,000-token limit attributed to prior models, potentially reducing how often unusually long generation tasks must be divided. Its benchmark position is more nuanced: Artificial Analysis places Argon high and GPT-6 Astra max at the same Intelligence Index level, while reporting substantially different hallucination rates. Buyers should therefore examine both configuration-specific quality measurements and the scheduled pricing increase when evaluating the model. What the sources don't address: how Argon’s benchmark and hallucination results translate into reliability across specific production workloads.