Apple M3 vs Apple M3 Ultra

The M3 Ultra delivers 9.3x the FP32 throughput of the M3 (33 vs 3.5 TFLOPS dense). The M3 Ultra also carries 488 GB more memory (512 GB vs 24 GB).

M3: Apple M3, 2023 M3 Ultra: Apple M3, 2025
FP32 dense lead
9.3x
M3 Ultra: 33 vs 3.5 TFLOPS
VRAM
24 vs 512 GB
488 GB more for the M3 Ultra
Bandwidth
100 GB/s vs 819 GB/s
M3 Ultra moves data faster

Specifications

M3
M3 Ultra
Architecture
Apple M3Apple M3
Launch Year
20232025
Form Factor
SoCSoC
VRAM
24 GB512 GB
Memory Bandwidth
100 GB/s819 GB/s
TDP
Process Node
3nm3nm

Performance (TFLOPS)

M3
M3 Ultra
FP32
3.5 TFlops 32.8 TFlops

FLOPS by Precision

What actually differs

The M3 Ultra is the newer part: Apple M3, launched in 2025, against the M3's Apple M3 from 2023. Newer architectures typically add lower-precision formats and better throughput per watt, so check the precision rows your workload actually uses.

Dense throughput for the M3 against the M3 Ultra: FP32 3.5 vs 33 TFLOPS. Sparse figures, where the vendor publishes them, appear under each dense number in the performance table.

Memory is 24 GB against 512 GB, fed at 100 GB/s versus 819 GB/s. More memory per GPU means larger models fit before you have to shard across cards, which often matters more than raw TFLOPS for inference.

For LLM training and inference, weight the FP8 and FP16 rows and memory capacity most heavily. For scientific and HPC workloads, the FP64 row is the one to read.

Frequently asked questions

Is the M3 faster than the M3 Ultra?

At FP32 precision the M3 Ultra reaches 33 TFLOPS dense against 3.5 TFLOPS for the M3. The performance table on this page lists every published precision for both GPUs.

Which has more memory, the M3 or the M3 Ultra?

The M3 Ultra carries 512 GB of VRAM versus 24 GB for the M3. Memory bandwidth is 100 GB/s for the M3 and 819 GB/s for the M3 Ultra.

Get Comparison Updates

New GPUs added weekly. Be the first to see how they compare.