UK The Short Answer
The T5000 and T4000 are the two modules in the Jetson Thor family, both built on the NVIDIA Blackwell GPU. The T5000 is the top variant, with up to 2,070 FP4 TFLOPS, 128 GB, and a 40 to 130 W envelope. The T4000 is the lower-power variant, with up to 1,200 FP4 TFLOPS, 64 GB, and a 40 to 70 W envelope. The T5000 is for the heaviest generative and physical AI; the T4000 trades peak performance for lower power and cost.
Where They Differ
The two share the Thor architecture, the connector, and the software, so the difference is performance and envelope. The T5000 carries roughly 1.7 times the AI compute and double the memory of the T4000, at up to nearly twice the power. The T4000 suits a workload that fits its budget where lower power and cost matter; the T5000 is for the largest models and the most concurrent pipelines.
-
//
T5000 Largest generative and physical-AI models, maximum memory and compute.
-
//
T4000 Lower power and cost, where the workload fits 64 GB and 1,200 TFLOPS.
-
//
Shared Same Blackwell architecture and software, so designs move between them.
Planning a Thor Build
Both Thor modules are a new generation with a new connector, power envelope, and thermal design, so a Thor system is a platform change rather than a drop-in. We can scope a T5000 or T4000 build as the platform reaches the market.