Navigating the current AI landscape often requires cross-referencing disparate data points, from reasoning scores on launch blogs to parameter counts buried in technical model cards. John Byron Hanby IV, founder and CEO of Iternal Technologies, developed the tool to consolidate this scattered information. Ultrabench tracks models by calculating a weighted average across ten core benchmarks, including GPQA Diamond and MMLU-Pro, while prioritizing independent results over vendor self-reported metrics. To qualify for the index, a model must be vetted by at least three distinct core benchmarks.
In section Releases
Iternal Technologies Launches Ultrabench to Simplify AI Model Selection
Austin-based Iternal Technologies has launched Ultrabench, a free aggregator that synthesizes 530 benchmarks into a single intelligence index for 309 large language models. By mapping performance scores alongside pricing, memory requirements, and hardware compatibility, the platform aims to replace fragmented vendor data with a unified, transparent resource for enterprise developers.
The platform provides a daily snapshot of the market, which as of September 30, saw Anthropic’s Claude Opus 5.5 leading with a 92.5 score and Moonshot AI’s Kimi K3 ranking highest among open-weights models at 84.4. Beyond intelligence, the service tracks price-per-million-tokens and hardware fit, specifying memory needs for BF16, INT8, and INT4 configurations. Hanby suggests the utility of the tool lies in its ability to help organizations identify the most cost-efficient models for specific tasks rather than defaulting to the most powerful option. The leaderboard is accessible without account registration, and the data is available via an open, keyless API designed for integration with autonomous AI agents.
Comments (0)
No comments yet. Be the first!