Data Center GPU Max 1350 holds 392 of the models in our catalogue and is, in practice, a IQ4_XS card — the largest it takes is dots.llm1.inst 142.8B at IQ4_XS.
Neighbours in memory rather than in price: memory decides whether a card can do the job at all, so two cards of the same size at different prices is the comparison you are making. Ordered by memory, then bandwidth — not by the value column, which is worked out from bandwidth and price alone and therefore rewards a cheap card whatever its software stack does to that bandwidth in practice. Read it as one input, not as a ranking.
Intel Data Center GPU Max 1350 — 96 GB VRAM.
- BRAND
- Intel
- VRAM
- 96 GB HBM2e
- BANDWIDTH
- 2460 GB/s
- FP16 COMPUTE
- 44.4 TFLOPS
- FP32 COMPUTE
- 44.4 TFLOPS
- TDP
- 450 W
- ARCHITECTURE
- Generation 12.5
With 96 GB VRAM and 2460 GB/s bandwidth, this GPU handles models up to 132B parameters.
Speed ≈ bandwidth / model_size × efficiency. A 7B model at Q4 runs at ~281 tok/s.
| MODEL | SIZE | VRAM Q4 | TOK/S | AVG |
|---|---|---|---|---|
| DBRX 132B | 132B | 81.2 GB | 22 | 46.3 |
| Mistral Medium 3.5 | 128B | 78.7 GB | 6 | 46.6 |
| Pixtral Large 124B | 124B | 76.3 GB | 6 | 39.3 |
| Nemotron 3 Super 120B-A12B | 123.61B | 76.0 GB | 66 | 53.2 |
| Mistral-Large 123B | 123B | 75.7 GB | 6 | 33.5 |
| Devstral 2 123B | 123B | 75.7 GB | 6 | 38.1 |
| Qwen 3.5 122B A10B | 122B | 75.1 GB | 79 | 56.8 |
| Nemotron 3 Super 120B | 120B | 73.8 GB | 66 | 57.3 |
| Mistral Small 4 119B | 119B | 73.2 GB | 121 | 50.2 |
| GPT-OSS 120B | 117B | 72.0 GB | 154 | 54.1 |
| Command A 111B | 111B | 68.3 GB | 7 | 27.6 |
| GLM 4.5 Air | 110B | 67.7 GB | 66 | 51.0 |
| Qwen 1.5 110B | 110B | 67.7 GB | 7 | 33.4 |
| Llama 4 Scout 17B-16E | 109B | 67.1 GB | 46 | 33.9 |
| Cogito v2 109B MoE | 109B | 67.1 GB | 46 | — |
| Ling 2.6 Flash | 107.49B | 66.2 GB | 106 | 36.8 |
| Sarvam 105B | 105B | 64.7 GB | 7 | 48.0 |
| Command-R+ 104B | 104B | 64.1 GB | 8 | 52.7 |
| Llama-3.2-90B-Vision-Instruct | 90B | 55.5 GB | 9 | 48.5 |
| Hunyuan A13B | 80B | 49.4 GB | 61 | 81.1 |