Gemini 4 Argon vs Claude Opus 5.5: Price and Benchmarks

In short

Use Gemini 4 Argon for long documents, video and knowledge work, and Claude Opus 5.5 for terminal-heavy coding agents. In Google's model-card table, Argon leads Opus 5.5 on AutomationBench (51.3% vs 42.5%), on GraphWalks at 256K to 1M tokens (84.2% vs 66.8%) and on LVBench video understanding (91.7% vs 83.7%), while Opus 5.5 leads on Terminal-bench 4.0 (66.4% vs 57.4%) and FrontierSWE v2 (62.3% vs 55.0%). Price matters less than it looks: Argon's API costs $2 per million input tokens and $10 per million output tokens until November 30, 2026, then $4 and $20 from December 1, which is Opus 5.5's price. In the apps, both need a paid plan: Google AI Pro at $19.99 a month or Claude Pro at $20.

On this page4

Argon costs half as much only until November 30

Per 1M tokens	Gemini 4 Argon, to Nov 30, 2026	Gemini 4 Argon, from Dec 1, 2026	Claude Opus 5.5
Input	$2	$4	$4
Output	$10	$20	$20
Cached input	$0.10	$0.20	$0.20 cache hit
Batch input / output	$1 / $5	$2 / $10	$2 / $10

Check prices on gemini.google

Compare the last two columns: from December 1 the rates match row for row, so any cost gap comes from how many tokens each model spends, not from the price list. Argon bills its thinking tokens as output at a medium thinking level by default, and Google does not say whether thinking can be switched off. Opus 5.5 always thinks, at medium effort by default; Anthropic says it no longer runs with thinking switched off. Neither charges more for long prompts: Argon has one rate across its 1,048,576-token window, and Anthropic includes Opus 5.5's 1M window at standard pricing.

Two billing differences remain. Anthropic charges to write a prompt cache ($5 per million tokens for five minutes, $8 for an hour), while Argon supports only implicit caching, with no write fee and no explicit cache control. Opus 5.5 also offers a fast mode at $8 input and $40 output per million tokens.

To make the dates concrete, take an illustrative job that reads 50 million input tokens and writes 5 million output tokens in a month: on Argon it costs $150 through November ($100 + $50) and $300 from December ($200 + $100), and on Opus 5.5 it costs $300 either way.

Where Google's benchmark table puts each model ahead

Only Google has published tables with both models in them. The one below comes from Argon's model card, which also lists GPT-6 Astra and Claude Fable 5.1; across the 18 rows where it gives both models a result, Argon leads 14 and Opus 5.5 leads 4. The figures are Google's.

Benchmark (Google's model card)	Gemini 4 Argon	Claude Opus 5.5
AutomationBench	51.3%	42.5%
Vals Finance Agent v2	65.4%	58.6%
GraphWalks 256K–1M BFS (F1)	84.2%	66.8%
LVBench	91.7%	83.7%
DeepSWE v1.1	77.9%	74.2%
FrontierSWE v2	55.0%	62.3%
Terminal-bench 4.0	57.4%	66.4%
Terminal-Bench Science 0.1	57.6%	63.3%
PostTrainBench	45.3%	49.3%

The table keeps all four of Opus 5.5's leads and Argon's widest ones. Read it in two halves: Argon's leads are on office and finance automation, very long inputs, video and one coding benchmark, DeepSWE; Opus 5.5's are on coding, science and ML-engineering tasks that run in a terminal. On Agent's Last Exam, a computer-use test, they are close (39.5% vs 38.2%).

Treat this as Google's measurement. Anthropic's own Opus 5.5 table does not include any Gemini model, and on three benchmarks with the same names it reports different Opus 5.5 results: 40.0% on AutomationBench where Google lists 42.5%, 58.7% on Terminal-Bench-Science 0.1 where Google lists 63.3%, and 89.0% on Chartography with tools where Google lists 66.3%. Terminal-Bench 4.0 is the one row where both report 66.4%. Google's Gemini product page has a second table, with some rows marked as Artificial Analysis runs, in which Opus 5.5 sits at 58 on the Artificial Analysis Intelligence Index and Argon at 50; it uses different settings, so its numbers are not mixed into the table above.

Argon can keep writing to 1M tokens; Opus 5.5 caps a request at 128K

Opus 5.5 returns up to 128K output tokens per request (up to 300K through the Message Batches API with a beta header). Argon's per-request cap is lower, 65,536 tokens on its model page (Google's guide to its latest model gives 32,768 for non-streaming requests), but its long-decode mode can keep generating up to 1M tokens minus the input. For a single very long output, such as a full translation of a long document, Argon is the one built for it.

Argon's knowledge cutoff is March 2026, Opus 5.5's June 2026. Argon accepts text, images, video, audio and PDFs and returns text only; its computer use is in preview.

Getting each model in the apps and the API

Gemini app: Google AI Pro ($19.99 a month) and Ultra ($99.99 or $199.99) include Argon; AI Plus gets the Flash model and free accounts Flash-Lite. Argon in the app is available in select countries.
Claude apps: Opus 5.5 is on Claude Pro ($20 a month, or $17 billed annually), Max (from $100), Team and Enterprise, not on Free.
API: Argon is gemini-4-argon, a stable model on the paid tier of the Gemini API in Google AI Studio, and Google Cloud offers it in its Gemini platform. Opus 5.5 is claude-opus-5-5 on the Claude API, Amazon Web Services, Google Cloud and Microsoft Azure.
Coding tools: Argon also runs in Google Antigravity, on AI Ultra by Google's launch post and on AI Pro and Ultra by the Antigravity blog.

Official sources

Editorial guidance grounded in official product sources.

Show all 12 sources

FAQ

Common questions

Is Gemini 4 Argon better than Claude Opus 5.5?

It depends on the work. In Google's model-card table Argon leads on knowledge work, long context and video (AutomationBench 51.3% vs 42.5%, LVBench 91.7% vs 83.7%), while Opus 5.5 leads on Terminal-bench 4.0 (66.4% vs 57.4%) and FrontierSWE v2 (62.3% vs 55.0%). Anthropic has published no comparison with Argon.

Is Gemini 4 Argon cheaper than Claude Opus 5.5?

Until November 30, 2026, yes: Argon costs $2 per 1M input tokens and $10 per 1M output tokens against Opus 5.5's $4 and $20. From December 1, 2026, Argon rises to $4 and $20, so the list prices match.

Is Gemini 4 Argon free?

No. Google says Argon is not available on the free tier of the Gemini API, and the free Gemini app uses Flash-Lite. You need paid API billing or a Google AI Pro ($19.99 a month) or Ultra plan.

Which has the bigger context window, Gemini 4 Argon or Opus 5.5?

Both read up to about 1M tokens with no long-prompt surcharge. They differ on output: Opus 5.5 writes up to 128K tokens per request, while Argon's default request cap is lower but its long decode can keep generating up to 1M tokens minus the input.

What is the Gemini 4 Argon API model ID and price?

The model ID is gemini-4-argon, a stable model available only on the paid tier of the Gemini API. It costs $2 per 1M input tokens and $10 per 1M output tokens, thinking tokens included, through November 30, 2026, then $4 and $20. Claude Opus 5.5 is claude-opus-5-5 at $4 and $20.

Next steps

Gemini 4 Argon vs Claude Opus 5.5: Price, Benchmarks and Which to Use