AI and Machine Learning
System76 AI machines combine the fastest computing, optimized thermals, and top-tier support — empowering professionals to accelerate their AI development.
Thelio AI WORKSTATIONS
Innovate faster with hardware and an operating system engineered to accelerate your AI development. AMD Threadripper CPUs and NVIDIA RTX Pro GPUs only perform as well as the system around them. Thelio workstations are engineered with optimized cooling to sustain peak performance, so you get every last core of performance from your investment. Fast hardware needs a fast operating system. System76’s Pop!_OS with the COSMIC Desktop was written from scratch in Rust for performance and security and designed with productivity features that make a noticeable difference in your day-to-day AI development work. Being based on Ubuntu, all of your AI tools are easily at hand. The System76 Thelio AI workstation lineup includes three models to align with your budget and AI workload requirements.
Thelio Mira AI
Thelio Mira AI prioritizes GPU compute with a 16-core CPU that requires less power, leaving more watts for dual, full-performance NVIDIA RTX Pro 6000 GPUs. The NVIDIA RTX Pro 6000 is 15%-30% faster per-gpu than the NVIDIA RTX Pro 6000 Max Q. For pure GPU-bound AI work like multi-gpu training, inference, fine-tuning, and image generation, Thelio Mira AI is an excellent choice that delivers maximum GPU performance per dollar.
Thelio Major
Thelio Major maximizes total system capability with up to 96 Threadripper CPU cores, faster PCIe lanes to the Dual NVIDIA RTX Pro 6000 Max Q GPUs, and more ECC system memory. AI workloads that require large dataset preprocessing, tokenization, decoding, and training pipelines with heavy CPU involvement benefit from Thelio Major’s combined CPU and GPU platform performance.
Thelio Mega
Thelio Mega is engineered for the largest local AI workloads with maximum compute and memory capacity. Mira AI and Major support up to 192 GB of GPU memory whereas Mega can support up to 288 GB. Look to Mega if your AI workload includes large-model fine-tuning and inference, multi-GPU distributed training, running multiple models simultaneously, or you require large datasets resident in system RAM.
Contact our experienced sales team if you need help deciding.
Thelio Mira AI
$3,299.00
Launching Sept. 9th!
Thelio Major
$6,999.00 $6,799.00
Thelio Mega
$8,499.00 $8,099.00
MOBILE AI WORKSTATIONS
System76 has you covered when your AI work needs to hit the road. Our mobile workstations come in two classes. An ultra-portable with combined system memory access for AI processing and models for NVIDIA GPU acceleration.
Pangolin Pro
With the Pangolin Pro you can run surprisingly large models on a thin-and-light laptop. The Pangolin Pro combines AMD Ryzen AI processing with up to 96GB of system memory that can be allocated to AI workloads in the BIOS. The Pangolin Pro is best for AI workloads that are constrained more by memory capacity than GPU throughput such as local LLM inference, coding assistants and local agents, and model experimentation.
Serval WS
The Serval WS hits the sweet spot for GPU-accelerated AI development and sub-5 lbs portability. With 12 GB of GDDR7 on a 192-bit bus providing 672 GB/s of memory bandwidth, the Serval WS can quickly move model weights, activations, and other data between VRAM and GPU compute cores. The Serval WS is your pick for small and medium model LLM inference, image generation, computer vision, CUDA development, and GPU accelerated data science.
Bonobo WS
The Bonobo WS is built for the heaviest AI workloads you’d want to run on a laptop. The NVIDIA RTX 5090 with 24 GB of VRAM doubles the Serval WS’s 12 GB and opens up much larger models with more demanding training workloads. Combined with up to 192 GB of system RAM for preprocessing, caching, and CPU offload, the Bonobo WS is excellent for LLM inference with larger models and longer context lengths, generative video, and multiple simultaneous AI workloads. It’s hefty at 8.6 lbs. but brings the power to take heavy AI workloads wherever you need to go.
Contact our experienced sales team if you need help deciding.
Pangolin Pro 16"
$1,699.00
Serval WS 16"
$3,599.00 $3,399.00
Bonobo WS 18"
$4,899.00
AI SERVERS
Build your own AI cluster with System76 deep learning servers and Ubuntu. Start with a single GPU server and scale out as compute demand grows. Each server and cluster is customized to your specific needs with up to eight NVIDIA RTX Pro 6000 Max Q GPUs for 768 GB of VRAM, AMD Epyc or Intel Xeon processors, and up to 4 TB of ECC system memory.
Pop!_OS Delivers
Pop!_OS is Built for your AI stack
Need some Guidance
Our sales team can help you find the perfect workstation, laptop, or server to accelerate your AI development. If you're an educational institution, you may qualify for discounts on desktops and servers with NVIDIA GPUs.