Tinygrad Launched Redesigned Tinybox AI Workstation
The company has updated its G4 system to support four AMD Radeon RX 9070 XT GPUs for deep-learning workloads.
Updated on Oct. 13, 2026 in Computers
Live Poll
Is now a good time for you to invest in high-performance workstation hardware?
Tinygrad has released a redesigned G4 workstation that features four AMD Radeon RX 9070 XT graphics cards. This custom-built machine is intended for deep-learning applications and is available for purchase now.
Why it matters
The system provides an accessible hardware path for demanding deep-learning workloads by prioritizing thermal management and acoustic output. This release offers a performance-focused alternative to enterprise-grade server hardware for developers.
The G4 workstation features a 32-core AMD EPYC processor, 32 GB of DDR4 memory, and a 1 TB SSD powered by a 1,700 W supply. The unit weighs 42 pounds and maintains noise levels below 50 dB using Noctua fans and liquid cooling.
The players
Tinygrad
A boutique hardware manufacturer specializing in custom workstations designed for artificial intelligence and deep-learning software development.
AMD
A semiconductor company that produces the EPYC processor and Radeon RX 9070 XT graphics cards used in the Tinybox architecture.
The details
Tinygrad builds these workstations to order, utilizing a specialized enclosure that increases physical spacing between graphics cards to facilitate improved airflow. This design allows for higher component density without exceeding the specified 50 dB noise threshold. The company differentiates this G4 model from its upcoming G5 platform, which is designed to support faster DDR5 memory and PCIe 5.0 connectivity.
Timeline
The Tinybox workstation release was officially announced on October 13, 2026.
Estimated delivery times for the custom-built workstations range between one and eight weeks after purchase.
The Tech Race
The G4 system serves as the current performance tier in Tinygrad’s catalog, functioning as a lower-cost, DDR4-based predecessor to the upcoming, higher-spec G5 platform. The G5 is expected to introduce advancements in memory bandwidth via DDR5 and data transfer speeds through PCIe 5.0.
The G4 workstation is currently available for order, with lead times estimated at one to eight weeks. Developers can choose between the G4 platform or wait for the future G5 model, which introduces DDR5 and PCIe 5.0 compatibility.
The takeaway
The G4 provides an immediate, air-cooled option for AI developers who need workstation-grade power without the noise of enterprise servers. Prospective buyers should track the pending release of the G5 platform to determine if the jump to PCIe 5.0 is necessary for their specific model training pipeline.
Further reading
Explore more hardware developments in our Computers section.
Source note: This article includes information reported by TechRadar.
Live Poll
Is now a good time for you to invest in high-performance workstation hardware?
