The 4 October CompareLLM note on Gemini 4 Argon described the 30 September announcement and restricted Fairwind access. That announcement is still the release date we store. What changed is access.
On 28 October the model author's Google AI Studio endpoint lists Argon at $2 input and $10 output per million tokens, with cache reads at $0.10. Context is 1,048,576 tokens. Median throughput was 41 tokens per second and median time to first token was 5.48 seconds, from OpenRouter's trailing thirty-minute window on that route.
The endpoint accepts text, image, audio, video and files, and it takes a reasoning-effort parameter.
Index, not a category exam
Artificial Analysis scores Argon 50 on Intelligence Index v5.0, configuration labelled high, read 28 October. GPT-6.1 Sol scores 55 on that index at the same $2/$10 token rates. Opus 5.5 scores 58 at $4/$20.
A v5.0 score does not fill SWE-bench, Arena, or our category recipes. Argon's category rows are labelled estimates with that index as the anchor.
The September announcement story stays as the record of restricted access. This note is the general-availability price.
