NVIDIA Nemotron Labs Dory1.0-4B
A downloadable English reasoning model with adjustable inference loops, long context, and local tool-calling APIs.
Gemini 4 Argon
ai.google.dev
Analyze multimodal inputs and generate text for complex coding, scientific reasoning, and tool-connected agent applications.
Visit website
Pricing
Paid
Available on
Gemini API, Google AI Studio
Category
Llm Models
What is Gemini 4 Argon?
Gemini 4 Argon is Google’s stable reasoning model for complex coding, mathematical and scientific problems, and multi-step agents. It accepts text, images, video, audio, and PDFs and generates text.
Available through the Gemini API and Google AI Studio, Argon offers adjustable thinking effort and long responses through background execution or GenerateContent client-side continuation.
Key features
Multimodal input
Combine written instructions with images, video, audio, and PDF documents.
Adjustable thinking effort
Choose low, medium, or high thinking. Medium is the default; higher effort increases token use and first-token latency.
Shared million-token context
Input, thinking, and output share 1,048,576 tokens. The default synchronous output limit is 65,536 tokens.
Two long-response options
Responses beyond 65,536 tokens can use background execution with a higher output limit or GenerateContent client-side continuation. Background execution manages continuation server-side. Both remain subject to the shared context and thinking budget.
Tools and structured output
Supports function calling, structured outputs, code execution, file search, search grounding, Google Maps grounding, and URL context.
Multiple inference options
Batch API, Flex inference, and Priority inference offer additional consumption options with separate prices.
Who is it for?
Software engineers reviewing complex changes
Analyze implementation material and develop proposed fixes for difficult coding tasks.
Researchers examining technical material
Combine documents and visual inputs to develop mathematical or scientific explanations for review.
Developers building multi-step agents
Connect reasoning to application functions and configured information tools.
How to use Gemini 4 Argon
- Open Google AI Studio and obtain a Gemini API key.
- Set up Cloud Billing for paid access to Argon.
- Use the Python or JavaScript SDK, or REST, and select gemini-4-argon.
- Submit your task and supported inputs, starting with medium thinking.
- For responses beyond 65,536 tokens, use background execution with a higher output limit or GenerateContent client-side continuation.
Gemini 4 Argon pricing
Paid
Argon is unavailable on the Free tier. API billing is token-based, including thinking tokens in output charges. Standard introductory rates apply through November 30, 2026, with higher rates from December 1. Grounding, Batch, Flex, and Priority have separate pricing.
Standard Paid Tier introductory rates
USD 2 per million input tokens; USD 10 per million output tokens
Metered token usage through November 30, 2026; no monthly subscription interval.
Output charges include thinking tokens. Implicitly cached input costs USD 0.10 per million tokens. Rates apply to Standard Paid Tier usage.
Standard Paid Tier from December 1, 2026
USD 4 per million input tokens; USD 20 per million output tokens
Announced metered rates starting December 1, 2026.
Output charges include thinking tokens. Implicitly cached input costs USD 0.20 per million tokens. Rates apply to Standard Paid Tier usage.
What to know before you use it
- Output is text only. Image generation, audio generation, and Live API are unsupported.
- Explicit context caching is unsupported; caching is implicit only. Temperature, top_p, and top_k settings are ignored.
- Thinking consumes the output budget, and input reduces the remaining shared context. Difficult requests may require longer timeouts.
- Computer use remains Preview despite the model’s stable designation.
Frequently asked questions
Can Argon generate images or spoken audio?
No. It accepts image and audio inputs but generates text.
Is background execution required for long responses?
No. GenerateContent also supports client-side continuation. Both options share the context window and thinking budget.
Is there a free API tier for Argon?
No. Gemini 4 Argon is unavailable on the Free tier and requires paid access.