Tenstorrent has unveiled TT-QuietBox 2, a liquid-cooled RISC-V AI workstation with a fully open source stack that brings teraflop-class inference and full stack control to the desktop.
Tenstorrent has unveiled TT-QuietBox™ 2 (Blackhole™), described as the industry’s first RISC-V AI workstation with a fully open-source software stack, bringing teraflop-class AI inference directly to the desktop. The liquid-cooled system is designed for quiet operation in labs and offices and will ship globally in Q2 2026, starting at $9,999.
The workstation runs with a fully open-source AI stack from compiler to kernel, enabling developers to inspect, modify, and control every layer of compute. Key components include TT-Forge, Tenstorrent’s open-source AI compiler, TT-Metalium, a low-level AI SDK with kernel-level control, and TT-LLK, the low-level kernel software. The stack provides full visibility across graph lowering, optimisation, execution, and hardware interaction, allowing developers to fork components, debug at the hardware level, and build custom AI pipelines.
The launch comes as AI inference now accounts for more than 55% of cloud AI infrastructure spending, valued at $37.5 billion globally, forcing developers to choose between cloud token fees or proprietary hardware stacks. QuietBox 2 positions itself as an alternative that enables local, fully transparent AI deployment.
“Tenstorrent is working hard on open source AI software and we wanted to build a teraflop development system that was easy to use in a lab or office, fast and quiet. It’s open top to bottom including the mechanical engineering. Build your own software or hardware. You can own your AI future,” said Jim Keller, CEO of Tenstorrent.
Built on RISC-V architecture, the system integrates four Blackhole ASICs delivering 480 Tensix cores and 2,654 TFLOPS at BlockFP8 precision, supported by 128GB GDDR6 and 256GB DDR5 memory.
QuietBox 2 can run large models locally, including GPT-OSS 120B, Llama 3.1 70B at 476.5 tokens per second, and Qwen3-32B for private coding workflows. Creative tools such as Flux for image generation and Wan 2.2 for video synthesis also run entirely on device.
The workstation runs Ubuntu 24.04, requires only a standard 120V wall outlet, and avoids rack infrastructure, making on-premise AI deployments accessible for developers, research labs, and small and medium businesses. Tenstorrent will demonstrate the system at Game Developers Conference (GDC) 2026.
