Agentic WeeklyGet the weekly

Gemini 4 Argon vs Claude Opus 5.5: API price before and after Dec 1

Gemini 4 Argon costs $2 per million input tokens and $10 per million output tokens through November 30, 2026, half of Claude Opus 5.5's $4 and $20. From December 1 Argon moves to $4 and $20, the same list price as Opus 5.5. Token counts per task differ by model. Last verified 2026-10-27 (America/Chicago).

By Agentic Weekly, operated independently from Texas, US · Last updated (America/Chicago) · How we work

Evidence: Reported. Prices and specs are read from Google's and Anthropic's pages listed under Sources on 2026-10-27 (America/Chicago). Benchmark rows are third-party mirrors with different harnesses. We have not run either model for this page.

Gemini 4 Argon vs Claude Opus 5.5 at a glance

Gemini 4 Argon and Claude Opus 5.5, read from Google's and Anthropic's pages on 2026-10-27
FeatureGemini 4 ArgonClaude Opus 5.5
Released2026-10-13 (Google launch post)2026-09-22 (Anthropic)
API model idgemini-4-argonclaude-opus-5-5
Input price, per 1M tokens$2.00 through 2026-11-30; $4.00 from 2026-12-01$4.00
Output price, per 1M tokens$10.00 through 2026-11-30; $20.00 from 2026-12-01 (includes thinking tokens)$20.00
Cached input, per 1M tokens$0.10, then $0.20 from 2026-12-01 (implicit caching only)$0.20 (cache read)
Batch input / output, per 1M tokens$1.00 / $5.00, then $2.00 / $10.00 from 2026-12-01$2.00 / $10.00
Context window1,048,576 tokens (input and output combined)1M tokens
Max output65,536 (default); up to 1,048,576 with long decode128K tokens
Input typesText, image, video, audio, PDFText, image
OutputTextText
Free tierNot available on the Gemini API Free tierNot covered on the pages we read; see Opus 5.5: free or paid?

Sources: Google pricing, Argon model page, Anthropic models overview, Anthropic pricing. Full facts: Gemini 4 Argon and Claude Opus 5.5.

Cost of one example task, before and after Dec 1

Illustration only, our arithmetic from list prices: one task with 200,000 uncached input tokens and 20,000 output tokens, with the same counts for both models. Real counts differ by model: Anthropic says Claude 4.7 and later models use a tokenizer that produces about 30% more tokens for the same text than earlier Claude models, and Google does not publish a like-for-like figure. Batch and caching change the bill.

Illustrative cost of one task (200K input + 20K output tokens) at standard list prices, our arithmetic, 2026-10-27
Model and periodList price per 1M tokensCost per taskCost per 100 tasks
Gemini 4 Argon, through 2026-11-30$2.00 in + $10.00 out$0.60$60.00
Gemini 4 Argon, from 2026-12-01$4.00 in + $20.00 out$1.20$120.00
Claude Opus 5.5$4.00 in + $20.00 out$1.20$120.00

At these counts, Argon is half the cost of Opus 5.5 until November 30 and the same cost from December 1. Neither figure is a measured bill. Our pricing guide has the full table: Gemini 4 Argon API pricing.

If you buy Argon through Google Cloud

Google's Cloud pricing page for Gemini on its agent platform shows Argon's promotional price as "50% credits back" on net spend through November 30, 2026, not a lower list price. It also says that when a Gemini 4 completion exceeds 32K tokens, users are charged for retrievable token usage even if the request fails on a client or server timeout, effective November 13, 2026. Read that page before budgeting for long completions: Google Cloud pricing. The Gemini API pricing page we cite above does not carry that note.

Reported benchmark rows, with the harness caveat

Both vendors make claims about their own model: Google says Argon is #1 on the Vals Index, and Anthropic says Opus 5.5 leads in agentic coding and knowledge work. We did not verify either. The only head-to-head rows we found are third-party mirrors of Terminal-Bench 4.0 on BenchLM, and each row uses its own harness and settings, so they are not a controlled comparison.

Terminal-Bench 4.0 rows mirrored by BenchLM, read 2026-10-27; settings differ by row
Source of the rowClaude Opus 5.5Gemini 4 Argon
Vals AI (Vals-hosted table)65.15%57.58% (high reasoning)
Artificial Analysis (its own harness)59.6%57.1%

Opus 5.5 scores higher than Argon in both tables, by 7.57 points and 2.5 points (our subtraction), but the gap changes with the harness, which is the point of the Terminal-Bench 4.0 explainer. BenchLM's official-leaderboard table, which lists Opus 5.5 at 64.85%, did not include an Argon row when we read it. Source: BenchLM Terminal-Bench 4.0.

Which to pick when

  • Budget before Dec 1 and the work can finish by Nov 30: Argon's list price is half of Opus 5.5's until then. Google gives no way to lock the introductory price past Nov 30.
  • Budget for December onward: the two have the same list price, so choose on fit, token use per task and tooling, not on the sticker.
  • Video, audio or PDF input: Argon accepts them per Google's model page; Opus 5.5's overview lists text and image input.
  • Very long outputs: Argon lists up to 1,048,576 output tokens with long decode; Opus 5.5 lists 128K max output.
  • Claude-specific tools or Anthropic's platform features: Opus 5.5 is Anthropic's recommended starting model for most workloads per its models overview.

What we did not test

We did not call either API or run either model, so we have no measurements of speed, accuracy, token use or cost on a real workload. Every figure is a vendor page or a cited third-party table read on 2026-10-27.

FAQ

Is Gemini 4 Argon cheaper than Claude Opus 5.5?

Until November 30, 2026, yes at list price: Argon is $2 input and $10 output per million tokens versus $4 and $20 for Opus 5.5. From December 1, Argon is $4 and $20, so the list prices match. Token use per task can still differ.

When does the Argon price change?

Google's pricing page says the introductory prices apply through November 30, 2026 and the higher prices start December 1, 2026: $4.00 input and $20.00 output per million tokens on the paid tier at standard rates. Argon is not available on the Free tier.

Which model scores higher on Terminal-Bench 4.0?

Opus 5.5 scores higher than Argon in the two third-party tables we found, by 7.57 and 2.5 points. The rows use different harnesses and settings, so treat them as reported figures, not a controlled head-to-head.

Does Argon support more input types than Opus 5.5?

Per vendor pages, yes: Google lists text, image, video, audio and PDF input for Argon, while Anthropic's overview says its current models support text and image input with text output.

Have you tested Gemini 4 Argon or Claude Opus 5.5?

No. This page reports vendor pages and third-party tables as read on 2026-10-27. We have not called either API or measured quality, speed or a real bill, so the cost example is arithmetic on list prices, not a measured cost.

Sources

Page history

  • : First published from Google's and Anthropic's pricing and model pages and BenchLM's Terminal-Bench 4.0 mirror, read on this date. Nothing was run. Planned: re-read on 2026-12-01.

Markdown version for agents: /compare/gemini-4-argon-vs-claude-opus-5-5.md