On Thursday, Google unveiled Gemini 4 Hydrogen, positioning the model as an accessible alternative for developers and everyday users seeking faster inference at a fraction of the price. The offering marks a shift in the company's strategy, making the model available to all Gemini app users—including those on free tiers—as well as through Google Search. Developers can access Hydrogen for $1.50 per million input tokens and $7.50 per million output tokens, representing a significant reduction compared to the pricing of Google's flagship Gemini 4 Argon.
The model processes up to 1 million input tokens simultaneously and completes tasks 2.5 times faster than the older Gemini 3.8 Flash. According to Google's benchmarking, Hydrogen achieves performance metrics that nearly rival Argon, with independent testing firm Artificial Analysis ranking it one point higher than Argon at default settings. Vals AI has identified Hydrogen as the cheapest and fastest option among its top six models.
Google designed Hydrogen to handle routine operations while reserving more demanding tasks for larger models. The company simultaneously introduced Orchestrator within its Antigravity coding tool, which intelligently routes requests between Hydrogen and Gemini 4 Carbon, Google's most powerful offering. This routing mechanism achieves results approaching Carbon's quality while consuming 24% less quota.
Reliability improvements accompany the new model's release. Hydrogen generates factual inaccuracies five times less frequently than Gemini 3.8 Flash. Within Google Search, the model addresses complex, multi-layered queries and surfaces links from a broader spectrum of sources.
Despite these advances, Hydrogen faces competition in the market. Artificial Analysis places its top-performing configuration one point below Anthropic's Claude Sonnet 5.5. OpenAI has also intensified competition by recently reducing prices on GPT-6 and introducing more economical model variants.
Source: The Next Web
