Gemini 4 Argon vs Claude Opus 5.5: specs and API pricing
Unofficial. Not affiliated with either provider.
Sources checked:
Model A is the first model named in the title and Model B is the second. The body of this page describes them neutrally and uses only what each provider publishes. Anything a provider does not publish is marked as not published.
Official specifications
| Dimension | Model A | Model B |
|---|---|---|
| Status | Stable. Public availability announced 13 Oct 2026. | Listed as the latest model. Released 22 Sep 2026. |
| Context window | 1,048,576 tokens, shared between input and output | 1M tokens |
| Maximum output | 65,536 tokens by default for synchronous requests; up to 1,048,576 with the documented long-decode protocol | 128K tokens; 300K on the batch API with a beta header |
| Input types | Text, image, video, audio and PDF | Text and images |
| Output types | Text | Text |
| Function calling | Supported | Tool use is referenced on the overview page; the full tool list is not given there |
| Structured outputs | Supported | Not stated on the overview page |
| Thinking | Thinking with low, medium and high levels | Adaptive thinking, always on; default effort is medium |
| Knowledge cutoff | Not published on the page checked | Jun 2026 (reliable cutoff) |
| API price per million tokens | Input $2 and output $10 (including thinking) through 30 Nov 2026; $4 and $20 from 1 Dec 2026. No free tier. | Input $4 and output $20. Batch requests 50% off; cache reads $0.20. |
| Where to use it | Google AI Studio and Google’s consumer app | The provider’s own API, Amazon Bedrock, Google Cloud, Microsoft Foundry and a platform on AWS |
| Independent benchmark score | Not published | Not published |
Model A: official specifications · Model A: official pricing · Model A: release post
Model B: official overview · Model B: official pricing · Model B: announcement
How to read the differences
- Start with input and output formats: check that the model accepts the files you plan to send.
- Then tools and integrations: use the providers’ API documentation for function calling and structured output.
- Then cost: use only the official pricing pages, and note that one price changes on a fixed date.
- Treat provider benchmarks as context only; their methods differ and cannot be subtracted from each other.
Not published, not used
Neither provider page we checked publishes a like-for-like benchmark of the two models, so this page cites no benchmark scores, latency figures or third-party leaderboards. We have not called either API or run our own tests.
FAQ
Which of the two models is stronger?
This page declares no winner. We have not run benchmarks, and neither official page we checked publishes a like-for-like comparison. Match the specification rows to your own task instead.
What are the context windows and prices?
See the table above, where each row links to the provider’s own page. One of the two prices changes on 1 Dec 2026.
How often is this page updated?
Only when an official page changes. The sources were last checked on 20 Oct 2026.