AI Agent Store/Model comparisonsBuild an AI worker ↗Google · Public API available
Use Argon for demanding coding, research and tasks with several steps. Public API access began October 13. Choose low, medium or high thinking; medium is the default.
Announced 2026-09-30 · Catalog added 2026-10-13 UTC · Checked 2026-10-15. Catalog addition and provider launch are separate dates.
We still need comparable independent tests. You can compare the confirmed features and limits below, but we cannot yet give this version an Agent Fit or Practical Value rating. Test it on your own tasks before switching; results for another version or thinking setting may not apply.
gemini-4-argon-high1525 ± 9 from 4,892 votes · Preliminary.
Source updated 2026-10-08; verified 2026-10-13.
Preference evidence for this exact configuration. It is not a percent-correct score or an effort sweep. This row is not merged with a differently configured LiveBench result.
Read the original user-preference result ↗Token prices measure the small pieces of text the model reads or produces. Capacity is the amount the service accepts in a request; check the limits below for any fallback to another model. A dash means the rate is unverified.
Exact catalog ID: google/gemini-4-argon. Reads text, images, video, audio and PDFs and returns text. It supports tools, structured answers and search; your application supplies permissions and runs its own tools.
Available thinking settings: low, medium, high. We do not yet have comparable independent results for these settings. The documented API default is medium when effort is omitted. This tells you what the API will use; it does not tell you which setting performs best.
Available through the paid Gemini API; no Free tier or Google AI Pro access. Caching is automatic only: explicit cache storage is not supported. The specialized cyber version still requires Fairwind approval; ordinary Argon access does not unlock it.
Long decode continues in steps of up to 64K tokens. Every step rereads earlier input and output, so billed input can exceed the final document size. Repeated input is eligible for implicit caching; use the response usage totals to estimate cost.
Input, thinking and answer share 1,048,576 tokens. Standard synchronous Interactions API output is capped at 65,536. Longer answers need background execution or continuation; the advertised 1M is not a single ordinary request allowance.
Standard launch prices through November 30, 2026 are $2 input, $0.10 cached input and $10 output per million tokens. From December 1 they become $4/$0.20/$20. Thinking is billed as output. Batch and Flex have separate half-price rates; Priority costs more. The calculator uses the standard routed offer and excludes continuation overhead and search fees. An invite-only Contributor tier offers lower prices in exchange for allowing Google to improve its products with the data. Those are not the standard prices shown here. Your bill can also include tools, retries and charged thinking tokens. Check the offer used by your application.
Provider release and specifications ↗
Inspect the exact catalog model ↗
More setup and licensing details ↗
More setup and licensing details ↗
More setup and licensing details ↗
Give each model the same examples, tools and definition of success. Record correct outcomes, response time, retries and total cost. A cheaper token price does not guarantee a cheaper completed task.
Compare caching and long-context costs ↗