NVIDIABlackwell Ultra

Best AI Models for NVIDIA B300 (288.0GB)

VRAM:288.0 GB HBM3e·Bandwidth:8000.0 GB/s·CUDA Cores:20,480·TDP:1400W

Datacenter Blackwell Ultra — rack/cloud only, not a desktop card. Listed for comparison.

With 288 GB of memory, this is a high-end configuration for local AI. You can comfortably run most open-source LLMs including large 70B parameter models at good quantization levels, making it one of the best setups for serious local AI work.

At this memory tier, nearly every popular open-source model is within reach. You can run Llama 3 70B at Q4_K_M or even Q5_K_M quantization with room to spare, handle coding assistants like DeepSeek Coder 33B at high quality, and easily run any 7B–30B model at full or near-full precision. Context windows remain generous even with larger models, so multi-turn conversations and long-document processing work smoothly.

Runs Well

  • 70B models (Llama 3 70B, Qwen 72B) at Q4–Q5
  • 30B models at Q6–Q8 quality
  • 7B–14B models at full FP16 precision
  • Vision models (LLaVA, CogVLM) without compromise

Challenging

  • Mixture-of-experts models like Mixtral 8x22B at higher quants
  • 120B+ models still require lower quantizations

What LLMs Can NVIDIA B300 Run?

138 models · 2 excellent · 4 good

Showing compatibility for NVIDIA B300

LLM models compatible with NVIDIA B300 — ranked by performance
ModelVRAMGrade
Qwen3 8B8.2B
Q4_K_M·942.0 t/s tok/s·41K ctx·EASY RUN
5.5 GBD26
GLM 4.6V107.7B
Q4_K_M·79.9 t/s tok/s·131K ctx·EASY RUN
65.1 GBC38
Q4_K_M·981.1 t/s tok/s·131K ctx·EASY RUN
5.3 GBD26
Q4_K_M·105.7 t/s tok/s·262K ctx·EASY RUN
49.2 GBC34
Q4_K_M·277.8 t/s tok/s·262K ctx·EASY RUN
18.7 GBD29
Q4_K_M·977.4 t/s tok/s·131K ctx·EASY RUN
5.3 GBD26
Gemma 3 27B IT27.4B
Q4_K_M·287.1 t/s tok/s·131K ctx·EASY RUN
18.1 GBD28
Gemma 3 12B IT12.2B
Q4_K_M·646.8 t/s tok/s·33K ctx·EASY RUN
8.0 GBD27
Q4_K_M·1793.1 t/s tok/s·262K ctx·EASY RUN
2.9 GBD26
Q4_K_M·6341.5 t/s tok/s·131K ctx·EASY RUN
0.8 GBD25
Q4_K_M·253.7 t/s tok/s·131K ctx·EASY RUN
20.5 GBD29
Qwen2.5 72B72.7B
Q4_K_M·116.6 t/s tok/s·131K ctx·EASY RUN
44.6 GBC33
Q4_K_M·1056.9 t/s tok/s·33K ctx·EASY RUN
4.9 GBD26
Q2_K·19.8 t/s tok/s·262K ctx·FAIR FIT
262.1 GBB48
Q4_K_M·1516.0 t/s tok/s·131K ctx·EASY RUN
3.4 GBD26
Q4_K_M·116.6 t/s tok/s·33K ctx·EASY RUN
44.6 GBC33

NVIDIA B300 Specifications

Brand
NVIDIA
Architecture
Blackwell Ultra
Compute Capability
10.0 (CUDA SM version)
VRAM
288.0 GB HBM3e
Memory Bandwidth
8000.0 GB/s
CUDA Cores
20,480
Tensor Cores
640
TDP
1400W
Release Date
2025-08-01

Get Started

Ollama (Recommended)

$curl -fsSL https://ollama.com/install.sh | sh
$ollama run llama3:8b

LM Studio

LM Studio

Download LM Studio, search for a model, and run it with one click.

Frequently Asked Questions

Can NVIDIA B300 run GLM 4.5?

Yes, the NVIDIA B300 with 288 GB can run GLM 4.5, GLM 4.6, Llama 3.2 90B Vision Instruct, and 1454 other models. 4 models run at excellent quality, and 19 at good quality. Check the compatibility table above for the full list with VRAM usage and estimated speed.

Is NVIDIA B300 good for AI?

The NVIDIA B300 has 288 GB of HBM3e, making it excellent for running local AI models. It supports 23 models at good quality or better. With 8000.0 GB/s memory bandwidth, it delivers fast token generation speeds. This is an enthusiast-grade GPU that handles most popular open-source LLMs.

How many parameters can NVIDIA B300 handle?

With 288 GB, the NVIDIA B300 supports models from 3B to 70B+ parameters depending on quantization level. At Q4_K_M (the recommended sweet spot), you can fit roughly 480B parameters. This means 7B models at high quality (Q6/Q8) or 30B+ models at Q4.

What quantization should I use on NVIDIA B300?

For the best balance of quality and speed on the NVIDIA B300, start with Q4_K_M — it preserves ~85% of the original model quality while keeping VRAM usage reasonable. With 24+ GB, you have the headroom to run 7B models at Q5_K_M or even Q6_K for noticeably better output quality. For larger 30B models, Q4_K_M remains the sweet spot.

How fast is NVIDIA B300 for AI inference?

With 8000.0 GB/s memory bandwidth, the NVIDIA B300 achieves approximately 1156 tokens/sec on a 7B model at Q4_K_M — that's very fast, well above conversational speed. A 14B model runs at ~578 tok/s. Token generation speed scales inversely with model size — smaller models are significantly faster.

tok/s = (8000 GB/s ÷ model GB) × efficiency

Smaller models = faster inference. Memory bandwidth is the main bottleneck for token generation speed.

Estimated speed on NVIDIA B300

Real-world results typically within ±20%. Speed depends on quantization kernel, batch size, and software stack.

Learn more about tok/s estimation →

What's the best model for NVIDIA B300?

The top-rated models for the NVIDIA B300 are GLM 4.5, GLM 4.6, Llama 3.2 90B Vision Instruct. The best choice depends on your use case: coding assistants benefit from code-tuned models, while general chat works well with instruction-tuned models like Llama or Qwen.

What power supply and cooling does NVIDIA B300 need?

The NVIDIA B300 has a TDP of 1400 W. A good rule of thumb is to provide at least double the GPU's TDP to cover the rest of the system — that means a 1600 W PSU or larger. At this power level, a high-airflow case matters: aim for at least two front intake fans and one rear exhaust, with tidy cabling so hot air isn't trapped around the card. LLM inference sustains full GPU load continuously — longer and more consistently than most gaming workloads — so also make sure your CPU cooler can keep up under combined load.

Anything to watch out for with NVIDIA B300?

Datacenter Blackwell Ultra — rack/cloud only, not a desktop card. Listed for comparison.