Every open model in our directory, ranked by how well it runs on SuperMicro Super AI Station (748GB) at 4-bit. Grades weigh whether the model fits in memory and how fast it should run.
SuperMicro Super AI Station has 748 GB of VRAM, enough to run 167 of the 169 open models we track at 4-bit. The largest that still fits well is LongCat-2.0 (1600B), and the top-rated is Llama 4 Maverick at about 39 tok/s.
167 of 169 models run comfortably on SuperMicro Super AI Station.
See full SuperMicro Super AI Station specs and pricingThe top models this device can run at 4-bit, ranked by fit and speed.
| Model | Grade | Speed | VRAM |
|---|---|---|---|
| Llama 4 MaverickMeta | SS | 39.1 tok/s | 146.4 GB |
| Llama 3.1 70B InstructMeta | SS | 50.7 tok/s | 112.8 GB |
| Llama 3.3 70B InstructMeta | SS | 50.7 tok/s | 112.8 GB |
| DeepSeek-V4-FlashDeepSeek | SS | 51.0 tok/s | 112.0 GB |
| Nvidia Nemotron 3 Super |
Llama 4 Maverick is the highest-rated model that runs on SuperMicro Super AI Station at 4-bit, at about 39 tok/s. In total, SuperMicro Super AI Station can run 167 of the 169 open models we track.
We size each model at 4-bit (Q4_K_M), add the KV cache for its context length and a fixed runtime overhead, then compare that to this device VRAM. Each model is graded on how comfortably it fits and how fast it should run on this card.
It is an estimate of decode speed in tokens per second at 4-bit, based on this device memory bandwidth and each model size. Real speed depends on your runtime, batch size, and context length, so use it to compare models, not as a hard promise.
Lower-bit quantization shrinks a model so more of them fit. These grades already assume 4-bit, which is the common local default. Going below 4-bit can squeeze a larger model in at some quality cost; going above needs more VRAM and may not fit.
| SS |
| 55.2 tok/s |
| 103.5 GB |
| GLM-5Z.ai | SS | 65.2 tok/s | 87.7 GB |
| GLM-5.1Z.ai | SS | 65.2 tok/s | 87.7 GB |
| Kimi K2.7 CodeMoonshot AI | SS | 66.3 tok/s | 86.2 GB |
| Kimi K2.6Moonshot AI | SS | 66.3 tok/s | 86.2 GB |
| Kimi K2 Instruct 0905Moonshot AI | SS | 67.6 tok/s | 84.6 GB |
| Kimi K2 ThinkingMoonshot AI | SS | 67.6 tok/s | 84.6 GB |
| Kimi K2.5Moonshot AI | SS | 67.6 tok/s | 84.6 GB |
Llama 4 MaverickMeta | 400B(17B active) | SS | 39.1 tok/s | 146.4 GB | |
| 70B | SS | 50.7 tok/s | 112.8 GB | ||
| 70B | SS | 50.7 tok/s | 112.8 GB | ||
DeepSeek-V4-FlashDeepSeek | 284B(13B active) | SS | 51.0 tok/s | 112.0 GB | |
Nvidia Nemotron 3 SuperNVIDIA | 120B(12B active) | SS | 55.2 tok/s | 103.5 GB | |
GLM-5Z.ai | 744B(40B active) | SS | 65.2 tok/s | 87.7 GB | |
GLM-5.1Z.ai | 744B(40B active) | SS | 65.2 tok/s | 87.7 GB | |
Kimi K2.7 CodeMoonshot AI | 1000B(32B active) | SS | 66.3 tok/s | 86.2 GB | |
Ad CodeRabbitAI-powered Code Reviews. Cut review time & bugs in half, instantly.Try for Free | |||||
Kimi K2.6Moonshot AI | 1000B(32B active) | SS | 66.3 tok/s | 86.2 GB | |
Kimi K2 Instruct 0905Moonshot AI | 1000B(32B active) | SS | 67.6 tok/s | 84.6 GB | |
Kimi K2 ThinkingMoonshot AI | 1000B(32B active) | SS | 67.6 tok/s | 84.6 GB | |
Kimi K2.5Moonshot AI | 1000B(32B active) | SS | 67.6 tok/s | 84.6 GB | |
GLM-4.6Z.ai | 355B(32B active) | SS | 81.3 tok/s | 70.3 GB | |
Mistral Large 3 675BMistral AI | 675B(41B active) | SS | 86.3 tok/s | 66.3 GB | |
DeepSeek-V3DeepSeek | 671B(37B active) | SS | 95.5 tok/s | 59.8 GB | |
DeepSeek-R1DeepSeek | 671B(37B active) | SS | 95.5 tok/s | 59.8 GB | |
Ad PayPerQPay-per-query access to top LLMs without a subscription. Use any model on demand.Try PayPerQ | |||||
DeepSeek-V3.1DeepSeek | 671B(37B active) | SS | 95.5 tok/s | 59.8 GB | |
DeepSeek-V3.2DeepSeek | 685B(37B active) | SS | 95.5 tok/s | 59.8 GB | |
Hy3Tencent | 295B(21B active) | SS | 102.6 tok/s | 55.7 GB | |
GLM-4.5Z.ai | 355B(32B active) | SS | 110.3 tok/s | 51.8 GB | |
GLM-4.7Z.ai | 358B(32B active) | SS | 108.6 tok/s | 52.6 GB | |
Kimi K2 InstructMoonshot AI | 1000B(32B active) | SS | 110.3 tok/s | 51.8 GB | |
| 70B | SS | 125.1 tok/s | 45.7 GB | ||
Qwen3.5-397B-A17BAlibaba | 397B(17B active) | SS | 124.2 tok/s | 46.0 GB | |
Ad Vast.aiAffordable on-demand GPU rentals for training and inference. Pick from thousands of hosts.Rent a GPU | |||||
Llama 2 70B ChatMeta | 70B | SS | 131.7 tok/s | 43.4 GB | |
Model Compatibility
Use the calculator to weigh this device against any other, model by model, with speed and fit scores.
Mixture-of-experts models only activate a fraction of their parameters at a time, so their runtime memory is closer to the active size than the headline size. We score fit on the active parameters, which is why some very large models still fit.
