Head-to-head comparison
Gemini 4 ArgonvsGPT-6.1 Sol
Gemini 4 Argon and GPT-6.1 Sol arrived within a day of each other at the same introductory price, $2 per million input tokens and $10 per million output, and they aim at different buyers. Argon is Google's frontier model for very long agent runs, gated to vetted defenders first. Sol is OpenAI's efficiency model, available immediately in the API, Codex, and ChatGPT Work.
At a glance
| Gemini 4 Argon | GPT-6.1 Sol | |
|---|---|---|
| Category | Chat | Chat |
| Function | Frontier model | Efficiency model |
| Pricing | Paid | Paid |
| Starting price | Introductory $2 / $10 per million tokens (standard $4 / $20); cached input $0.10 | $2 / $10 per million tokens; cached input $0.10 |
| Released | September 2026 | September 2026 |
Gemini 4 Argon
Pros
- Output limit an order of magnitude above most frontier models — suits very long agent runs
- Introductory price matches the $2/$10 tier set by Sonnet 5.5 and GPT-6.1 Sol
- Multimodal input in one model
Cons
- Not generally available at launch — gated to the Fairwind Program with no public date
- Introductory price doubles after an unannounced promotional period
- Benchmarks are Google-reported and unverified independently
GPT-6.1 Sol
Pros
- Brings GPT-6-class work down to the $2/$10 price tier
- Spares plan usage limits compared with Astra in ChatGPT Work
- Immediately available across OpenAI surfaces
Cons
- Not available in the standard ChatGPT chat picker at launch — Work and Codex only
- "Near-Astra" is OpenAI's claim; independent evaluations pending
- Arrived alongside a cut to the $200 Pro plan's allowance
Which one should you pick?
Pick Gemini 4 Argon if
You need a 1M-token output limit for long-running coding, legal, or finance agents, you build on Google Cloud, or you qualify for the Fairwind Program and need cyber-defence capability now.
See Gemini 4 Argon details →Pick GPT-6.1 Sol if
You want GPT-6-class work at the $2/$10 tier today, you already run Codex or ChatGPT Work, or you need a model your whole team can access without a waitlist.
See GPT-6.1 Sol details →Bottom line
Sol is the one you can use this week. Argon is the one to plan for if your agents run for hours. At the same price, availability decides it for most teams until Google opens Argon up.