Best local AI models for AMD FirePro M6000

The largest models that fit fully

The 30 largest of the 56 models that fit; every smaller model in the catalog fits too. Best quant means the highest quality compression whose weights and 4k context both sit inside the memory.

Close, but only with CPU offload

These need more than the card holds at their smallest practical quant, so part of the model runs from system memory (figures assume 32 GB of it). They work, several times slower.

How to read this

The AMD FirePro M6000 is an older mobile workstation graphics card equipped with 2 GB of GDDR5 video memory. This memory size is the absolute limit for running local AI models entirely on the hardware. To fit within this 2 GB frame, models must use quantization. Quantization is a compression method that reduces the precision of model weights to save space. The quant column shows the best format that fits your hardware. For example, a Q4_K_M quant uses four bit quantization, while Q8_0 uses eight bit precision.

Several small models can run completely inside the 2 GB video memory. The Allegro 2.8B model fits at a Q4_K_M quant using exactly 2 GB. The Open-Sora Plan 2.7B model also fits at Q4_K_M using 2 GB. You can run LFM2 2.6B or Playground v2.5 at Q4_K_M using 1.9 GB. Stable Diffusion 3.5 Medium and Canary 2.5B both fit at Q4_K_M using 1.8 GB. SeamlessM4T v2 2.3B and Kimi K3 DSpark 2.2B fit at Q5_K_M using 2 GB. Parler-TTS 2.2B fits at Q5_K_M using 1.9 GB.

Smaller models can run at higher precision levels. SmolVLM 2B, Stable Diffusion 3 Medium, Pyramid Flow, and Wav2Vec2 XLS-R all fit at Q6_K using 2 GB. Moondream 2 fits at Q6_K using 1.9 GB. Qwen3 1.7B and SmolLM2 1.7B fit at Q6_K using 1.7 GB. StableLM 2 1.6B, Sana 1.6B, Zonos 0.1, and Dia 1.6B all fit at Q8_0 using 2 GB. Whisper Large v3 fits at Q8_0 using 2 GB. ControlNet, Hunyuan-DiT, Stable Video Diffusion, Whisper Large v2, AudioGen, and AudioLDM 2 all fit at Q8_0 using 1.9 GB. Tango 2 fits at Q8_0 using 1.8 GB. TinyLlama 1.1B and SantaCoder 1.1B fit at Q8_0 using 1.4 GB.

When a model is too large for the 2 GB video memory, you must use CPU offload. This process splits the model between your graphics card and your system RAM. Offload allows you to run larger models, but it costs speed. Your system will run much slower because system RAM is slower than GDDR5 video memory. We assume a system with 32 GB of system RAM for these offload cases.

For offload, SmolLM3 3B, Replit Code v1.5 3B, Kandinsky 3.1, Voxtral Mini, Orpheus TTS, and Higgs Audio v2 need 2.2 GB at Q4_K_M and 4.2 GB of system RAM. MusicGen needs 2.4 GB at Q4_K_M and 4.4 GB of system RAM. Stable Diffusion XL needs 4.1 GB at FP8 and 6.1 GB of system RAM. SDXL Turbo and SDXL Lightning both need 2.6 GB at Q4_K_M and 4.6 GB of system RAM.

You must remember the 4k context caveat when running these models. The memory numbers listed only cover the model weights. Running a model with a long text history or a large context window requires extra video memory. If you use a full 4k context window, the system will need more memory than the base model size, which might trigger slow CPU offload.