Artificial Analysis' Intelligence Index: Gemini 4 Hydrogen nearly ties Claude Sonnet 5.5 and GPT-6.1 Sol, at a third of Sonnet's cost but ~2.5x that of Sol
First reported by X ·
Advanced AI models now offer comparable intelligence at varying price points, affecting developer tool choices.
Google's new Gemini 4 Hydrogen model has achieved an intelligence score close to that of Anthropic's Claude Sonnet 5.5 and OpenAI's GPT-6.1 Sol, according to an analysis by Artificial Analysis. While Hydrogen's performance is near the top tier, its cost is approximately one-third of Claude Sonnet 5.5's but 2.5 times that of GPT-6.1 Sol. The Gemini 4 Hydrogen, the latest release from Google DeepMind in a series of new models, features an "xhigh" reasoning level. The model demonstrated a notable characteristic by honestly reporting its own errors, such as a "minor spill" and an "unclosed lid" during testing, indicating a level of self-awareness in its performance reporting. This positions Hydrogen as a competitive offering in the advanced AI model market, balancing performance with a distinct cost structure.
The emergence of Gemini 4 Hydrogen at a cost significantly lower than Claude Sonnet 5.5, while nearly matching its intelligence, indicates a market trend towards more accessible high-performance AI. This development challenges established pricing strategies and suggests that the premium for top-tier AI capabilities may be diminishing. Developers and businesses can now expect a broader range of cost-effective options for sophisticated AI tasks, potentially accelerating adoption across industries that previously found cutting-edge models prohibitively expensive.
This competitive pricing, especially relative to GPT-6.1 Sol, signals an intensified race in the LLM market where performance parity is being achieved with more aggressive cost-saving measures. Google's strategy with Gemini 4 Hydrogen suggests a focus on democratizing access to advanced AI. The model's self-reporting honesty on errors, while a qualitative observation, could also imply a move towards more transparent and reliable AI development, encouraging greater trust and adoption among users evaluating these powerful new tools.
AI-written summary. May contain errors.