Gemini 4 Argon
Overview
Gemini 4 Argon is a Google DeepMind reasoning model announced on September 30, 2026, and generally available since October 13, 2026. It is designed for complex software engineering, scientific reasoning, and enterprise knowledge work, with a unified 1M-token context window and long-decode generation.
Key Features
- Adjustable Reasoning: Supports
low,medium(default), andhighthinking levels to balance reasoning depth, latency, and token use. - Agentic Coding: Designed for demanding debugging, code migration, and iterative engineering workflows.
- Enterprise Analysis: Supports reasoning across documents and evidence for complex professional knowledge work.
- Multimodal Understanding: Accepts text, image, video, audio, and PDF input, including charts and long videos, and produces text output.
- Defensive Cybersecurity: Can identify, validate, and patch software vulnerabilities in authorized agent workflows.
- Extended Generation: Long decode supports generation beyond 64K tokens within the remaining unified context budget, subject to interface and execution requirements.
Best Use Cases
- Software Maintenance: Multi-step debugging, code migration, and optimization requiring iterative analysis and verification.
- Professional Research: Financial investigation and legal research or drafting that combine evidence across documents.
- Visual Knowledge Work: Chart interpretation and video analysis alongside document-based reasoning.
- Security Remediation: Authorized vulnerability discovery and patch validation, subject to access controls and safeguards.
Capabilities and Limitations
| Capability | Description |
|---|---|
| Reasoning | Dynamic thinking with low, medium, and high levels; medium is the default. |
| Multimodal | Text, image, video, audio, and PDF input; text output. |
| Context Window | 1,048,576 tokens shared across input, thinking, and output. This is a unified budget, not separate input and output allowances. |
| Max Output | 65,536 tokens for synchronous requests. Provider long decode can use up to 1,048,576 minus input tokens, including thinking, where supported by the interface. |
| Tool Use | Provider capabilities include function calling, structured outputs, code execution, File Search, URL context, Search and Maps grounding, and Computer Use (Preview). B.AI availability depends on the endpoint. |
| Video Understanding | Agentic video understanding is generally available; video processing defaults to static. |
| Prompt Caching | Implicit caching only, enabled by default for prompts of at least 4,096 tokens. Explicit caching is unsupported. |
| Knowledge Cutoff | March 2026; knowledge in some domains may be less current. |
Known Limitations
- Output Budget: Thinking consumes the generation budget. A small limit can produce truncated or empty answers; longer input leaves less room for generation.
- Long Decode: On the provider's Interactions API, outputs above 65,536 tokens require background execution and a raised output limit. Do not assume the same behavior is available through every B.AI endpoint. Exact word-count or token-length targets are unreliable.
- Unsupported Features: Explicit caching, native image or audio generation, and the Live API are unsupported. Computer Use remains a preview capability.
- Sampling Controls:
temperature,top_p, andtop_kare ignored. - Reliability and Latency: Hallucinations, occasional slowness, and timeouts remain possible; difficult prompts can take several minutes.
- Evaluation Conditions: Provider evaluations may use the highest thinking settings and different test environments. Evaluation scores do not guarantee production outcomes.
Standard Pricing
The token prices below are shown in USD per 1 million tokens; web search is billed in USD per use.
| Model | Input (USD / 1M Tokens) | Cache Write (USD / 1M Tokens) | Cache Read (USD / 1M Tokens) | Output (USD / 1M Tokens) | Web Search (USD / use) |
|---|---|---|---|---|---|
| Gemini 4 Argon | $2.00 | $2.00 | $0.10 | $10.00 | $0.014 |
Credits settlement: B.AI converts charges at 1 USD = 1,000,000 Credits and deducts Credits from the account balance.
Web search costs 14,000 Credits per use, equivalent to $0.014 per use.
This model supports implicit caching only. The cache-write price shown in the table does not imply support for explicit cache creation. For provider implicit caching, prompts must meet the minimum length of 4,096 tokens; meeting the length threshold does not guarantee a cache hit on every request.
Prices shown in the documentation are B.AI standard reference prices for base billing purposes. B.AI may provide lower actual usage costs through account benefits. Specific prices and account benefits are subject to the platform display and final billing records.