Clarifai and Vultr have announced impressive new benchmark results showcasing exceptional speed, efficiency, and scalability in AI inference on GPUs. At the NVIDIA GTC conference in Washington, D.C., the two companies highlighted their collaboration, demonstrating the capabilities of the Clarifai Reasoning Engine on Vultr’s extensive GPU clusters. Independent tests revealed that the engine processes 544 tokens per second and achieves a time to first token of just 0.36 seconds, with a cost efficiency of $0.16 per million tokens on the GPT-OSS-120B model. This performance surpasses other GPU-based platforms and approaches the efficiency of specialized ASIC accelerators.
Kevin Cochrane, CMO of Vultr, emphasized the synergy between software innovation and cloud engineering, which enables rapid development of high-performance AI solutions. Matthew Zeiler, CEO of Clarifai, highlighted Vultr’s infrastructure as crucial for maximizing the potential of the Clarifai Reasoning Engine, ensuring performance and cost efficiency without compromising quality.
The recent benchmarks coincide with the Clarifai 11.9 release, which includes new cloud instances on NVIDIA’s advanced hardware and expanded toolkit compatibility. The platform’s continuous optimization for enterprise-scale workloads allows for improved performance over time.
Clarifai and Vultr are setting new standards in AI inference, offering developers and enterprises the tools to drive innovation in reasoning, agentic systems, and generative AI.
