Google's New AI Gemini 4 Argon Takes on OpenAI's GPT-6 Astra

In September, OpenAI launched GPT-6 Astra, and Google followed with Gemini 4 Argon on September 30. Both are among the most powerful AI models available, designed to handle coding, complex tasks, and professional work. But according to independent analysis from Artificial Analysis, they score equally on overall intelligence, so neither is clearly 'smarter' than the other.

The biggest difference is price. During its launch promotion, Argon costs $2 per million input tokens and $10 per million output tokens, while Astra costs $10 and $50 respectively. Even after Argon's promotion ends (likely rising to $4 and $20), it remains less than half the price of Astra. However, Astra uses fewer tokens per task, which can narrow the cost gap. One report estimates Argon at $1.99 per task during the promotion vs Astra at $3.26, rising to $3.98 later.

Availability also differs: Astra is available now through ChatGPT Work, Codex, and the API, though enterprise access is off by default. Argon is initially limited to vetted cyber defenders in Google's Fairwind program, with no public API date announced. Both models have a 1-million-token context window (enough to process about 700,000 words at once), but Google hasn't confirmed Argon's input limit.

In terms of capabilities, Astra offers five reasoning effort settings, but at the highest setting, it's slow—taking over 4 minutes to start answering. It also provides only brief summaries of its reasoning, making it harder to audit. Both companies restrict cyber capabilities, refusing offensive security tasks. For most users, the decision boils down to budget and timing: Argon for cost savings (once available), Astra for immediate use and stronger math/agent scores.