Meta MTIA 450 vs Meta MTIA 100
The MTIA 450 delivers 68.4x the BF16 throughput of the MTIA 100 (3,500 vs 51 TFLOPS dense). The MTIA 450 also carries 224 GB more memory (288 GB vs 64 GB).
Meta MTIA 450
Full specs →Meta MTIA 100
Full specs →Specifications
Performance (TFLOPS)
FLOPS by Precision
What actually differs
The MTIA 450 is the newer part: MTIA, launched in 2027, against the MTIA 100's MTIA from 2023. Newer architectures typically add lower-precision formats and better throughput per watt, so check the precision rows your workload actually uses.
Memory is 288 GB against 64 GB, fed at 18.4 TB/s versus 176 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 MTIA 450 faster than the MTIA 100?
At BF16 precision the MTIA 450 reaches 3,500 TFLOPS dense against 51 TFLOPS for the MTIA 100. The performance table on this page lists every published precision for both GPUs.
Which has more memory, the MTIA 450 or the MTIA 100?
The MTIA 450 carries 288 GB of memory versus 64 GB for the MTIA 100. Memory bandwidth is 18.4 TB/s for the MTIA 450 and 176 GB/s for the MTIA 100.
How much power do the MTIA 450 and the MTIA 100 draw?
The MTIA 450 is rated at 1.4 kW TDP and the MTIA 100 at 25 W.
Get Comparison Updates
New GPUs added weekly. Be the first to see how they compare.