Gemini 4 Argon
Beta ⧉Gemini 4 Argon Beta supports streaming text chat and Low, Medium and High thinking levels; Low is the default. Thinking tokens are billed as output. With web search on, the model looks things up with Google Search and cites its sources; each search costs $0.0035. The official context is about 1 million tokens shared by input and output, and it passed our long-input retrieval test. Repeated long inputs have produced implicit cache hits. Only actual hits receive the cached rate; a hit is not guaranteed on every repeat.
Why this model
Gemini 4 Argon Beta: three thinking levels, Google Search and a shared million-token context
Highlights
- Beta · reasoning · Google Search
- Low / Medium / High
- About 1M-token context · shared input and output
- Implicit cache hits observed on repeated long inputs. Actual hits cost $0.05 per million tokens; each repeat may differ.
Price vs. official list
Per 1M tokens, USD · You save ↓70%
Use cases
Live Test · Playground
Try out Gemini 4 Argon right here (available after login).
Input
Advanced
Conversation
About Gemini 4 Argon
Gemini 4 Argon Beta supports streaming text chat and Low, Medium and High thinking levels; Low is the default. Thinking tokens are billed as output. With web search on, the model looks things up with Google Search and cites its sources; each search costs $0.0035. The official context is about 1 million tokens shared by input and output, and it passed our long-input retrieval test. Repeated long inputs have produced implicit cache hits. Only actual hits receive the cached rate; a hit is not guaranteed on every repeat.
FAQ
How do I turn on web search?
How do I control thinking?
What is the context window?
Do repeated requests hit the cache?
What does Beta cover?
Related Models
Explore other models you can integrate.
GPT-5.6 Sol
Frontier flagship for agentic coding
GPT-5.6 Luna
Fast, economical agentic coding model
GPT-6 Astra
Next-generation OpenAI flagship - 1.05M context
GPT-6.1 Sol
The GPT-6 Sol upgrade - near-Astra coding and agents
GPT-6 Sol
Built for complex coding and agentic workflows - 1.05M context
GPT-6 Luna
The most efficient GPT-6, built for high-volume work
GPT Image 2
High-quality text-to-image / image-to-image model
GPT Image 2.5 Flare
Next-gen image model - clean and smooth
GPT Image 2.5 Sunburst
Next-gen image model - richer texture
Nano Banana 2
High-quality text-to-image / image-to-image model
Nano Banana Pro
Flagship text-to-image / image-to-image model
Nano Banana 2.1
The upgraded Nano Banana 2, with better images and text
Grok Imagine Image
xAI's image model · renders in seconds
Grok Imagine Image Quality
xAI's higher-quality tier · finer detail and lettering
Claude Sonnet 4.6
Balanced, efficient chat and coding model
Claude Opus 5
Next-generation flagship from Anthropic
Claude Fable 5
Anthropic Fable series · narrative and long-form writing
Claude Opus 5.5
Anthropic's next Opus · 1M context · adaptive thinking
Claude Sonnet 5.5
Anthropic's newest Sonnet · the best mix of speed and intelligence
Gemini 3.1 Pro
Next-generation flagship reasoning model from Google
Gemini 3.8 Flash
Latest Flash · adaptive thinking
DeepSeek V4.1 Flash
DeepSeek's current Flash model - thinking on or off
DeepSeek V4 Flash 0731
Pinned V4 Flash snapshot - a version that stays put
GLM-5.3
The new Z.ai flagship - coding and agents
Qwen3.8 Max
Alibaba's Qwen 3.8 flagship - reasoning, coding and agents
Qwen3.8 Flash
The light, fast Qwen3.8 - high volume, low cost
Grok 4.7
xAI's flagship reasoning model - long context
Veo 3.1
Text / image / first-last frame / references · 8 s
Gemini Omni Flash
Google's next-generation video model · 6 / 8 / 10 s
Seedance 2.5 · Token
Built for reliable delivery and fast creation
Seedance 2.5 · Token Draft
Draft to final
Seedance 2.5
Text / image to video · up to 30 seconds
Seedance 2.0
Text / image to video · 5-15 seconds
Seedance 2.0 Fast
Low latency · quick turnaround
Seedance 2.0 Mini
Entry tier · lowest cost
Seedance 2.0 · Clip
Special price · per clip · 5 / 10 / 15 s at one price
Seedance 2.5 · Special
Temporary availability · per clip
Wan 3.0
One endpoint, five modes · up to 30 seconds
Wan 3.0 Prime
Same capabilities · several times faster
MiniMax H3
1080P · native audio
Grok Imagine Video 1.5
Billed per clip · up to 15 seconds
