Artificial Analysis benchmarks two Gemini 4 Argon variants
Artificial Analysis has published separate results for the High and Medium settings of Google's Gemini 4 Argon, showing that the higher-intelligence setting was also faster, while Medium cost slightly less per task.
As we reported earlier, Gemini 4 Argon matched GPT-6 Astra on Artificial Analysis's intelligence rankings. The benchmarking firm has now published a release page that splits the model into two reasoning variants, Gemini 4 Argon (High) and Gemini 4 Argon (Medium), and compares them on intelligence, speed and cost.
Here is how the two variants compare:
- Intelligence: Gemini 4 Argon (High) is the top model in the release, scoring 50 on the Intelligence Index. The page excerpt does not give a score for Medium.
- Speed: High took the least time per Intelligence Index task, at 30.1 minutes.
- Latency: High also had the lowest time to first answer token, at 42.66 seconds.
- Cost: Medium was the cheapest at $4.31 per task. Prices across the two variants differ by at most 1.2x.
Both variants are reasoning models. They accept text and image input, produce text output, and have a 1M-token context window.
Artificial Analysis also lists the evaluations behind its capability indexes. They include AA-Briefcase v1.1, AA-Omniscience v1.1, HLE-Diamond, AutomationBench-AA, GDP.pdf, MLCR-AA, CritPt v2 (Hard Subset), Terminal-Bench 4.0, Terminal-Bench-Science 0.1 and FrontierCode-AA 2.0. The page does not include each variant's per-index scores, Medium's speed and latency, or High's cost per task.
The small price gap matters for builders choosing a setting. On these numbers, Medium saves relatively little per task, and High finished tasks faster.