Gemini 4 Argon API Pricing 2026: $2/$10 Intro, $4/$20 Standard
Jonhisking · October 3, 2026 · Updated October 15, 2026

Google announced Gemini 4 Argon on September 30, 2026, its new top model. It is built for coding, cybersecurity and long agentic workflows, and Google says it sets a new state of the art on DeepSWE v1.1 (77.9%) and leads the Vals Index, without naming which rival models it was measured against. Here is what it costs on the API, what a real request costs, and how it compares with GPT and Claude.
Gemini 4 Argon API pricing

Prices per million tokens. It starts at an introductory rate and moves to the standard rate later.
| Rate | Input | Cached input | Output |
|---|---|---|---|
| Introductory | $2 | ~$0.10 | $10 |
| Standard | $4 | ~$0.20 | $20 |
Cached input is 95% off the normal input price, which matters a lot when you resend the same instructions or documents. The introductory rate runs until November 30, 2026; from December 1 the standard rate applies, according to Google's launch announcement and launch coverage. Check Google's pricing page for the rate in force.
Argon can also output up to 1 million tokens in a single response. Long outputs cost accordingly: 1 million output tokens is $20 per response at the standard rate.
Can you use it now?
Yes. Since October 13, 2026, Gemini 4 Argon has been generally available to developers through the Gemini API in Google AI Studio, and to Google AI Pro and Ultra subscribers in the Gemini app, per Google's announcement and 9to5Google. On the API it is paid only: reports say there is no free tier for Argon. The model ID is gemini-4-argon.
Updated October 15, 2026: when this post was first published on October 3, Argon was limited to partners in Google's Fairwind cyber-defense program.
How it compares

A typical request: 2,000 input tokens and 500 output tokens.
| Model | Per 1M tokens (in/out) | 1 request | 10,000 requests |
|---|---|---|---|
| Gemini 3.8 Flash | $0.75 / $3.75 | $0.0034 | $34 |
| Gemini 4 Argon (intro) | $2 / $10 | $0.009 | $90 |
| GPT-6 Sol | $2 / $10 | $0.009 | $90 |
| Claude Sonnet 5.5 | $2 / $10 | $0.009 | $90 |
| Gemini 3.1 Pro | $2 / $12 | $0.010 | $100 |
| Gemini 4 Argon (standard) | $4 / $20 | $0.018 | $180 |
| Claude Opus 5.5 | $4 / $20 | $0.018 | $180 |
| GPT-6 Astra | $10 / $50 | $0.045 | $450 |
| Claude Fable 5.1 | $10 / $50 | $0.045 | $450 |
Other prices are official API prices checked in October 2026.
What stands out:
- The standard price matches Claude Opus 5.5 exactly: $4 in, $20 out.
- The introductory price matches GPT-6 Sol and Claude Sonnet 5.5. A top model at a mid-tier price, for now.
- Even at the standard price it is 60% cheaper than GPT-6 Astra and Claude Fable 5.1. If Google's benchmark claims hold, it is the best value among the top models.
- It costs more than Gemini 3.1 Pro, about 80% more per request at the standard price. If 3.1 Pro already does the job, there is no reason to switch.
When to use it
Good fit: large code changes and migrations, security reviews, long multi-step agent runs, and analysis of long videos or charts. The more a mistake costs you, the more it is worth paying for.
Overkill: summaries, translation, classification and short answers. Gemini 3.8 Flash handles those at about 80% less cost per request.
A common setup is to send everyday requests to Flash or 3.1 Pro and route only the hard ones to Argon.
How to spend less

- Test during the introductory period. It is half the standard price, the best time to see if it fits your workload.
- Use caching. Cached input is 95% off, so keep fixed instructions and documents at the start of the prompt.
- Cap the output. It can write up to 1 million tokens, and output costs 400% more than input. Ask for only what you need.
- Write prompts in English. Most other languages take more tokens for the same text. The 💸 Save tokens button in the token counter rewrites a prompt in English on your own device.
Check your own cost
Gemini 4 Argon is now in the token counter at the standard price. Paste a prompt you actually use and compare Argon, GPT-6 and Claude side by side.
Prices and rollout dates can change. Check the Gemini API pricing page before large jobs.
Sources
- Google: Gemini 4 Argon announcement
- Gemini API pricing
- OpenAI API pricing
- Claude API pricing
- Google on X: Gemini 4 Argon availability and introductory price
- 9to5Google: Gemini 4 Argon rolling out (October 13, 2026)
- VentureBeat: Gemini 4 Argon broadly available
More articles
- Claude Haiku vs Opus for Coding: 330 Measured Runs
- Claude Code Cache Lasts 1 Hour: What a Break Costs (Measured)
- Write Shorter Prompts Without Losing Quality: Measured Before and After
- Why Output Tokens Cost More (and 6 Ways to Use Fewer)
- What Is a Token? A Plain-English Guide for AI Users
Paste a prompt to compare its cost on Gemini 4 Argon, GPT-6 and Claude.
Open the token counter →