NVIDIA Jetson AGX Thor Developer Kit (128GB)
Compare NVIDIA Jetson AGX Thor Developer Kit (128GB) against another machine in the Arena
Reference: Nvidia Jetson on Wikipedia
- Class
- SoC
- Memory
- 128 GB LPDDR5X
- Runs (Q4_K_M)
- —
- Bandwidth
- 273 GB/s
- TDP
- 130 W
- Released
- 2025-08-19
- Price (US)
- $5499 new as of 2026-07
Sourcing and disambiguation notes
Same GB/s bandwidth as DGX Spark (273 GB/s LPDDR5X, 256-bit bus) but the edge-robotics form factor and 130W envelope trade off against DGX Spark's ~1.05-1.5x higher token-generation throughput on the same llama.cpp build, per a direct head-to-head test. Vendor "2070 TOPS FP4" is a Blackwell sparse-FP4 compute claim, not an inference speed measurement -- not recorded as a benchmark here. 2026-07 update: price rose to $5,499 after NVIDIA's July 2026 Jetson lineup price increase (up to 101% on some SKUs, driven by memory costs), up from the $3,499 launch price (Aug 2025) previously recorded. Kit was reported out of stock direct from NVIDIA's US marketplace at research time; third-party resellers (Micro Center, Seeed Studio, Arrow) listed it. Apple and Jetson 'new' figures across this dataset postdate the 2026 memory-shortage price rise -- fresh only as of their as_of date.
Specification sources
- https://www.cnx-software.com/2025/08/19/3499-nvidia-jetson-agx-thor-developer-kit-2070-tops-jetson-t5000-som-for-robotics-and-edge-ai/ — press, 2025-08-19
Measurements
Decode is token generation — the speed you feel while an answer streams. Prefill is prompt processing — the wait before it starts. Why bandwidth predicts decode speed.
No records yet for this hardware. Know of a published benchmark on it? Submit the link.
What verified, single-source and estimated mean, and the same rows with every filter and sort in the benchmarks explorer.
Speed over time
What it can run
Fit is arithmetic, not a measurement — how it is computed. Against 128 GB; models that fit are listed largest first.
No modeled quant fits in 128 GB.