AMDCDNA 3

Best AI Models for AMD Instinct MI300X (192.0GB)

VRAM:192.0 GB HBM3·Bandwidth:5300.0 GB/s·Stream Processors:19,456·TDP:750W

With 192 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 AMD Instinct MI300X Run?

158 models · 124 excellent · 14 good

Showing compatibility for AMD Instinct MI300X

LLM models compatible with AMD Instinct MI300X — ranked by performance
ModelVRAMGrade
Qwen3 235B A22B235.1B
Q4_K_M·22.5 t/s tok/s·41K ctx·DECENT
141.6 GBB63
MiniMax M2228.7B
Q4_K_M·23.1 t/s tok/s·197K ctx·DECENT
137.8 GBB64
Q4_K_M·22.5 t/s tok/s·262K ctx·DECENT
141.6 GBB63
Q4_K_M·22.5 t/s tok/s·262K ctx·DECENT
141.6 GBB63
Step 3.7 Flash201.4B
Q4_K_M·23.9 t/s tok/s·262K ctx·DECENT
132.9 GBB64
DeepSeek v2235.7B
Q4_K_M·22.0 t/s tok/s·164K ctx·DECENT
144.3 GBB62
Q4_K_M·22.0 t/s tok/s·164K ctx·DECENT
144.3 GBB62
DeepSeek V2.5235.7B
Q4_K_M·22.0 t/s tok/s·164K ctx·DECENT
144.3 GBB62
Q4_K_M·24.2 t/s tok/s·200K ctx·DECENT
131.6 GBB64
Q4_K_M·32.0 t/s tok/s·1049K ctx·RUNS WELL
99.5 GBA71
GPT OSS 120B120.4B
Q4_K_M·43.8 t/s tok/s·131K ctx·RUNS WELL
72.7 GBA77
Q4_K_M·42.6 t/s tok/s·262K ctx·RUNS WELL
74.7 GBA76
Q4_K_M·44.4 t/s tok/s·RUNS WELL
71.7 GBA77
GLM 4.5 Air110.5B
Q4_K_M·47.7 t/s tok/s·131K ctx·RUNS WELL
66.7 GBA79
Q4_K_M·37.4 t/s tok/s·66K ctx·RUNS WELL
85.1 GBA74
Q4_K_M·37.4 t/s tok/s·66K ctx·RUNS WELL
85.1 GBA74

AMD Instinct MI300X Specifications

Brand
AMD
Architecture
CDNA 3
VRAM
192.0 GB HBM3
Memory Bandwidth
5300.0 GB/s
Stream Processors
19,456
FP16 Performance
1307.40 TFLOPS
TDP
750W
Release Date
2023-12-06

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.

Prompt Processing

Estimated for Step 3.5 Flash Base, the compute-bound phase that reads your prompt before the first reply token appears.

727.2tok/s prefill

Short chat

704 ms

512 tok prompt

Long chat

5.6 s

4,096 tok prompt

Document / codebase

45.1 s

32,768 tok prompt

Prefill is compute-bound and a different number from the decode tok/s shown elsewhere on this page — how prompt processing works →

Efficiency & Value

Based on Step 3.5 Flash Base at ~24.4 tok/s decode.

Tokens per watt

0.03tok/s per W

Higher is better.

How efficiency & value are calculated →

Performance figures are estimates calibrated as of 2026-07-30 see calibration basis →

GPUs to Consider Over AMD Instinct MI300X

Similar GPUs and upgrades with more VRAM or higher bandwidth for AI

Frequently Asked Questions

Can AMD Instinct MI300X run Qwen3 235B A22B?

Yes, the AMD Instinct MI300X with 192 GB can run Qwen3 235B A22B, MiniMax M2, Qwen3 235B A22B Instruct 2507, and 1744 other models. 1585 models run at excellent quality, and 108 at good quality. Check the compatibility table above for the full list with VRAM usage and estimated speed.

Is AMD Instinct MI300X good for AI?

The AMD Instinct MI300X has 192 GB of HBM3, making it excellent for running local AI models. It supports 1693 models at good quality or better. With 5300.0 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 AMD Instinct MI300X handle?

With 192 GB, the AMD Instinct MI300X supports models from 3B to 70B+ parameters depending on quantization level. At Q4_K_M (the recommended sweet spot), you can fit roughly 320B parameters. This means 7B models at high quality (Q6/Q8) or 30B+ models at Q4.

What quantization should I use on AMD Instinct MI300X?

For the best balance of quality and speed on the AMD Instinct MI300X, 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 AMD Instinct MI300X for AI inference?

With 5300.0 GB/s memory bandwidth, the AMD Instinct MI300X achieves approximately 707 tokens/sec on a 7B model at Q4_K_M — that's very fast, well above conversational speed. A 14B model runs at ~353 tok/s. Token generation speed scales inversely with model size — smaller models are significantly faster.

tok/s = (5300 GB/s ÷ model GB) × efficiency

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

Estimated speed on AMD Instinct MI300X

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 AMD Instinct MI300X?

The top-rated models for the AMD Instinct MI300X are Qwen3 235B A22B, MiniMax M2, Qwen3 235B A22B Instruct 2507. 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 AMD Instinct MI300X need?

The AMD Instinct MI300X has a TDP of 750 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.