AMD Instinct MI300X 192GB: Specs & Local-AI Compatibility
Class-leading 192GB HBM3 — big models on a single accelerator.
Specs
- Memory
- 192 GB
- Memory type
- HBM3
- Bandwidth
- 5,300 GB/s
- Approx FP16
- to verify
- Architecture
- CDNA 3
- Process
- TSMC 5nm/6nm
- Power
- 750 W
- Launch
- 2023
Models this chip can run
Open models graded for a single AMD Instinct MI300X 192GB, best fit first.
- Qwen3 235B-A22B (MoE)Qwen · ~235B · 128K ctx · Apache-2.0
Fits at Q4_K_M (~130GB) with ~39GB headroom — about 1 concurrent instance.
Q4_K_M · ~130GBRuns well - Qwen2.5 72BQwen · ~72B · 128K ctx · Qwen License
Fits at FP16 (~145GB) with ~24GB headroom — about 1 concurrent instance.
FP16 · ~145GBRuns well - Llama 3.1 70BLlama · ~70B · 128K ctx · Llama Community License
Fits at FP16 (~140GB) with ~29GB headroom — about 1 concurrent instance.
FP16 · ~140GBRuns well - Llama 3.3 70BLlama · ~70B · 128K ctx · Llama Community License
Fits at FP16 (~140GB) with ~29GB headroom — about 1 concurrent instance.
FP16 · ~140GBRuns well - DeepSeek-R1 Distill Llama 70BDeepSeek · ~70B · 128K ctx · MIT
Fits at FP16 (~140GB) with ~29GB headroom — about 1 concurrent instance.
FP16 · ~140GBRuns well - Mixtral 8x7B (MoE)Mistral · ~47B · 32K ctx · Apache-2.0
Fits at FP16 (~90GB) with ~79GB headroom — about 1 concurrent instance.
FP16 · ~90GBRuns well - CodeLlama 34BCodeLlama · ~34B · 16K ctx · Llama Community License
Fits at FP16 (~68GB) with ~101GB headroom — about 2 concurrent instances.
FP16 · ~68GBRuns well - Qwen2.5 32BQwen · ~32B · 128K ctx · Apache-2.0
Fits at FP16 (~64GB) with ~105GB headroom — about 2 concurrent instances.
FP16 · ~64GBRuns well
Devices built on this chip
Build a private AI Business OS on AMD Instinct MI300X 192GB
Run your own AI agents on hardware you control — private by design, no per-seat data leaving your premises. BrainOutput helps you pick the right machine and turn it into a working AI Business OS.
Explore the AI Business OS