Tiny Corp is now shipping specialized AI computers called “tinybox” designed for deep learning training and inference, according to the company’s website. The computers are available in two configurations currently shipping, with a third “exabox” model planned for 2027.
The company offers two models currently in stock: a “red v2” configuration priced at $12,000 featuring 4x AMD 9070XT GPUs with 64GB total GPU RAM and 778 TFLOPS of FP16 performance, and a “green v2” model at $65,000 with 4x RTX PRO 6000 GPUs providing 384GB GPU RAM and 3086 TFLOPS. Both systems run Ubuntu 24.04 and ship within one week of payment.
The hardware specifications show the green v2 model includes 32-core AMD GENOA processors, 192GB system RAM, 4TB RAID storage plus 1TB boot drive, and 2x 10GbE networking. The red v2 uses a 32-core AMD EPYC CPU with 128GB system RAM and 2TB NVMe storage. According to the specifications, the green v2 system measures 20x8x8.5 feet and weighs 20,000 pounds, requiring a concrete slab installation.
Tiny Corp states the tinybox “was benchmarked in MLPerf Training 4.0 vs computers that cost 10x as much” and describes it as “likely the best performance/$” for deep learning applications. The company emphasizes that “anything that can train can do inference,” positioning the systems for both training and deployment of AI models.
The planned “blackwell exabox” for 2027 would feature approximately 1 exaflop of performance with 720x RDNA5 AT0 XL GPUs, 25,920GB GPU RAM, and an estimated price of $10 million. Tiny Corp accepts only wire transfers for payment and offers no customization options to maintain “low prices and quality high.”
