Question 1

What AI models can run on Gaudi 3?

Accepted Answer

The Gaudi 3 (128 GB VRAM) can run models including gemma-4-E4B-it, gemma-4-26B-A4B-it, gemma-4-26B-A4B, gemma-3-12b-it, gemma-4-31B and more — 10 compatible models indexed so far. Compatible models are determined by whether the model's memory footprint fits within the GPU's 128 GB VRAM, potentially using quantization (Q8, Q6, or Q4) to reduce memory usage.

Question 2

What is the cheapest cloud option for Gaudi 3?

Accepted Answer

We don't currently have cloud pricing data for the Gaudi 3. This GPU may not be widely available from major cloud providers, or pricing data hasn't been synced yet.

Question 3

Is Gaudi 3 good for fine-tuning?

Accepted Answer

Yes, the Gaudi 3 is excellent for fine-tuning. With 128 GB of VRAM, it can handle full fine-tuning of models up to ~30B parameters and LoRA/QLoRA fine-tuning of much larger models. It's a top-tier choice for training workloads.

Question 4

How much VRAM does Gaudi 3 have?

Accepted Answer

The Gaudi 3 has 128 GB of VRAM (HBM2e). It also offers 3700 GB/s memory bandwidth, 1,835 TFLOPS of FP16 compute, FP8 support for efficient inference.

Question 5

What quantization formats work on Gaudi 3?

Accepted Answer

The Gaudi 3 supports the following quantization formats: FP16 (full precision), FP8, Q8 (8-bit), Q6 (6-bit), Q4 (4-bit). FP16 works for models up to ~40B parameters. FP8 halves memory vs FP16 with minimal quality loss. Q4 quantization lets models up to ~435B parameters (by VRAM estimate) fit on the Gaudi 3.

Gaudi 3

Specifications

Cloud pricing

Compatible models

Compare this GPU

Frequently asked questions

VRAM	128 GB
Memory type	HBM2e
Bandwidth	3700 GB/s
FP16 TFLOPS	1,835
FP8 support	Yes