Replacing the Flash series
Google is retiring its Flash series in favor of Gemini 4 Hydrogen. The new model is positioned as the smaller, faster tier, but independent benchmarks show it performing much closer to the pricier Gemini 4 Argon than expected.
What the independent index shows
According to the Artificial Analysis Intelligence Index, Hydrogen scores nearly level with Claude Sonnet 5.5 at a lower price. However, its cost per task is roughly 2.5 times that of GPT-6.1 Sol, meaning the model can still be expensive in real-world use despite its lower sticker price.
Behind on agentic coding
The tests also show Hydrogen still trails rivals on agent-based coding tasks. That suggests the model is strong for fast chat and quick-response work but weaker on multi-step autonomous coding.
Why it matters
Hydrogen getting this close to Argon shows Google is rebalancing price and performance across its model lineup. The fact that the larger Argon, launched just four weeks earlier, is already being challenged points to a shift where cost and task fit matter more than picking the biggest model.
