Rebellions logo

Rebellions RebelPOD (32-node)

pod
256× Rebel100
Backend fabric 400G QSFP112-DD; frontend network 25G SFP28; optional 100G storage network
FP16
262
PFLOPS
FP8
524
PFLOPS
Power
224.0
kW Total
Memory
36864
GB Total

The 32-node RebelPOD is eight NPU racks plus one management rack, holding 32 RebelServers and 256 Rebel100 cards. Rebellions rates it at 524.3 PFLOPS FP8 and 262.1 PFLOPS FP16, with 36.86 TB of HBM3E at 1,228.8 TB/s and 48 TB of DDR5 across 64 AMD EPYC 9355 processors totalling 2,048 cores and 4,096 threads. The NPU racks draw up to 224 kW and the whole pod is air-cooled, which is unusual at this scale. Rebellions also offers 64, 128 and 256-node configurations but publishes specifications only for this 32-node one, so the larger sizes are deliberately not listed here. All figures are dense, per Rebellions' own labelling of the Rebel100 compute; no sparse rate is published.

We will point you at suppliers who have it. Free, and no signup.

FP16
262
PFLOPS
FP8
524
PFLOPS

System Details

GPU Configuration

GPU Count: 256 GPUs
Architecture: REBEL
Interconnect: Backend fabric 400G QSFP112-DD; frontend network 25G SFP28; optional 100G storage network

System Specifications

Form Factor: pod
Total Power: 224.0 kW
Total Memory: 36.9 TB
Memory Bandwidth: 1.2 PB/s

Precision Performance Breakdown

PrecisionSystem PerformancePer GPUEfficiency
262 PFLOPS 1,024 TFLOPS —
524 PFLOPS 2,048 TFLOPS —

Where a vendor states a basis, the figure shown is the dense one, and any figure it publishes is listed separately. Vendors commonly headline the sparse number instead: for NVIDIA's 2:4 structured sparsity that is exactly twice the dense figure, though other vendors' sparse modes do not all follow that ratio.

Powered by Rebellions Rebel100

This system utilizes 256 × Rebellions Rebel100 GPUs, each delivering exceptional performance for AI and HPC workloads.

Per GPU TDP

600W

Per GPU Memory

144 GB

Process Node

4nm

Architecture

REBEL

Documentation & Resources

Flopper Spec Sheet

Rebellions RebelPOD (32-node) specifications, generated from our database. Printable.

Vendor product page

Rebellions RebelPOD (32-node) documentation

View ↗

ATOM-Max Server brochure v1.6 (Rebellions)

Rebellions • 2026-06-18

View Document ↗

Typical Use Cases

AI/ML Training
High-Performance Computing
Data Analytics

The Rebellions RebelPOD (32-node) runs 256× Rebel100 GPUs over Backend fabric 400G QSFP112-DD; frontend network 25G SFP28; optional 100G storage network, delivering 524 PFLOPS FP8 dense.

© 2026 Flopper.io - Compare the hardware powering AI