Google's Gemini 4 Hydrogen beats its larger sibling Argon in independent index
Four weeks after its launch, Google's Gemini 4 Hydrogen beats its larger sibling model Argon in the independent Artificial Analysis Intelligence Index and comes close to Claude Sonnet 5.5.
Google released Gemini 4 Hydrogen four weeks before this report, and the model already shows remarkable performance in independent tests. In the Artificial Analysis Intelligence Index, an independent benchmark, Hydrogen outperforms the larger model Argon from the same family. A comparatively smaller system therefore achieves better results than its own sibling.
Hydrogen replaces the previous Flash model series. In the index, it is nearly on par with Claude Sonnet 5.5, the model from competitor Anthropic. At the same time, Hydrogen is cheaper to operate than Claude Sonnet 5.5. For companies and developers who focus on performance and cost, this combination of results and price is a central point of the assessment.
The cost comparison with GPT-6.1 Sol, however, paints a different picture. Measured per task rather than per token, Hydrogen costs about 2.5 times as much according to the analysis. Anyone judging by the cost of individual completed work steps must therefore take the price gap to this competing model into account, even though Hydrogen performs well in other comparisons.
According to the available information, Hydrogen shows weaknesses in agentic programming tasks, where the model still lags behind competitors. It is also unclear whether Google will discontinue the larger Argon model in the future. According to the reports, the company has not yet issued any binding statement on this.
The assessment is based on one outlet's analysis of the benchmark data. The results refer to a specific index and do not measure every application equally well. Whether the results are confirmed in practical use will likely only become clear with further independent tests.
Source: the-decoder.com
53
