Model comparison
Gemini 4 Neon vs Gemini 4 Argon
Gemini 4 Neon costs $3.00 / $15.00 and Gemini 4 Argon $2.00 / $10.00 per 1M input / output tokens. Both accept about 1M tokens of context, text, images and files.
Which one to choose
Picked by a published rule, not by an editor. Scenarios without data are left out.
Everyday requests
Gemini 4 Argon
$0.011 vs $0.007 for 1,000 input and 500 output tokens.
Long prompts
Gemini 4 Argon
A 500K-token prompt with a 2K-token answer costs $1.53 on Gemini 4 Neon and $1.02 on Gemini 4 Argon.
Context window
About the same
Gemini 4 Neon: 1.05M tokens; Gemini 4 Argon: 1.05M. Differences under 10% count as a tie.
How the verdicts are calculated
- Everyday requests: lower cost for 1,000 input + 500 output tokens at standard rates.
- Long prompts: lower cost for a 500K-token prompt and 2K-token answer, including long-context surcharges.
- Context window: bigger window wins; differences under 10% count as a tie.
- Mixed inputs: more input types accepted; a tie is not shown until multimodal benchmark scores are collected.
Not shown yet
- Coding: needs SWE-bench Verified scores
- Speed: needs independent output-speed measurements
- Privacy: needs data-retention and training-use policies
Side by side
The better value in each row is marked. Observed from LiteLLM and OpenRouter, checked against vendor pricing pages.
| Spec | Gemini 4 Neon | Gemini 4 Argon |
|---|---|---|
| Input priceper 1M tokens | $3.00 | $2.00 (better) |
| Output priceper 1M tokens | $15.00 | $10.00 (better) |
| Cached inputper 1M tokens | $0.15 | $0.10 (better) |
| Typical request1,000 input + 500 output tokens | $0.011 | $0.007 (better) |
| Long-prompt pricingCompared on a 500K-token prompt | No surcharge listed | No surcharge listed (better) |
| Batch priceinput / output per 1M tokens | $1.50 / $7.50 | $1.00 / $5.00 (better) |
| Context window | 1.05M tokens | 1.05M tokens |
| Max output | 66K tokens | 66K tokens |
| Knowledge cutoff | Mar 2026 | Not disclosed |
| Input types | Text, Images, Video, Audio, Files (PDF) | Text, Images, Video, Audio, Files (PDF) |
| Reasoning | Adjustable effort | Adjustable effort |
| Tool use | Yes | Yes |
| Structured output (JSON) | Yes | Yes |
| BenchmarksGPQA Diamond, SWE-bench Verified, AIME 2025, LMArena | Not collected yet | Not collected yet |
| Output speedtokens per second | Not collected yet | Not collected yet |