Undominated.ai

Self-host

MI300X

192 GiB vendor-published framebuffer. Catalogue open-weight models that fit, ranked by measured quality. NVIDIA does not pick the winner.

192 GiB framebuffer. Source: https://www.amd.com/en/products/accelerators/instinct/mi300.html

14 of 25 profiled open-weight models fit 192 GiB (vram) at F16, 8192 context. Best rated that fits: Gemma 4 31B.

Weights 61.41 GiB KV not estimated Runtime 1 GiB Headroom 129.6 GiB / 192 GiB

mi300x: vendor 192 GiB framebuffer, not free-after-driver.

Among models that fit

Highest LMArena Elo among models that fit: Gemma 4 31B 1443.4. Elo is a human-preference scale, not the intelligence index.

The arena boards are separate scales and are never blended. A missing cell is unmeasured, not zero. A model’s highest axis is relative to its own scores, not a claim it leads the catalogue.

Rated models that fit

ModelHuman preferenceQuant that fitsEstimated GiBHeadroom
Gemma 4 31B
visionvideoreasoningtools
Fit assumptions
  • KV heads and head dim are not in the public size table we used. Cache is not estimated.
  • KV cache not estimated — layer/head geometry is not in the sourced profile. Longer context may not fit.
1443.4F1662.41129.59
Qwen3.8 27B
visionvideoreasoningtools
Fit assumptions
  • Hybrid: only 16 of 64 layers are Gated Attention. KV uses kvLayers=16, not nLayers=64. Measured Q4 is Unsloth UD-Q4_K_M, not a vanilla Q4_K_M.
1440.7F1655.5136.5
DeepSeek V4 Flash 0423
reasoningtools
Fit assumptions
  • Hybrid CSA + HCA. KV cache is not estimated with the dense formula. DeepSeek’s own PDF also writes Flash as 285B; we use the 284B table on the Hub card.
  • MoE: memory follows 284B total parameters, not 13B active.
  • KV cache not estimated — layer/head geometry is not in the sourced profile. Longer context may not fit.
1432.1Q4_K_M172.46519.535
gpt-oss-120b
reasoningtools
Fit assumptions
  • MoE: memory follows 117B total, not 5.1B active. Vendor 80GB claim is for MXFP4, not Q4_K_M. KV from the published config: 36 layers, 8 KV heads, head dim 64.
  • MoE: memory follows 117B total parameters, not 5.1B active.
1365.4Q8_0125.87566.125
Phi 4
Fit assumptions
  • Hub safetensors listing says 15B params; the card states 14B. We use the card. KV from config.json: 40 layers, 10 KV heads, hidden 5120 → head dim 128.
1216.7F1630.563161.438

Fits, unrated

Does not fit, or not profiled

  • z-ai/glm-5.3 — even Q4_K_M wants 455.624 GiB; pool is 192 GiB
  • qwen/qwen3.8-2.4t-a95b — even Q4_K_M wants 1450.719 GiB; pool is 192 GiB
  • moonshotai/kimi-k3 — even Q4_K_M wants 1691.5 GiB; pool is 192 GiB
  • z-ai/glm-5.2 — even Q4_K_M wants 455.624 GiB; pool is 192 GiB
  • moonshotai/kimi-k2.7-code — even Q4_K_M wants 604.75 GiB; pool is 192 GiB
  • nvidia/nemotron-3-ultra-550b-a55b — even Q4_K_M wants 333.063 GiB; pool is 192 GiB
  • minimax/minimax-m3 — even Q4_K_M wants 259.405 GiB; pool is 192 GiB
  • deepseek/deepseek-v4-pro — even Q4_K_M wants 967 GiB; pool is 192 GiB
  • mistralai/mistral-large-2512 — even Q4_K_M wants 408.531 GiB; pool is 192 GiB
  • deepseek/deepseek-v3.2 — even Q4_K_M wants 406.116 GiB; pool is 192 GiB
  • meta-llama/llama-4-maverick — even Q4_K_M wants 242.5 GiB; pool is 192 GiB

How the number is made

Weights: a measured GGUF size when we have one, otherwise total parameters × bits per weight ÷ 8. Q4_K_M is treated as 4.83 bits/param. MoE memory uses total parameters, not active parameters.

KV cache, when layer and head geometry is sourced: 2 × kvLayers × kvHeads × headDim × context × 2 bytes (FP16, batch 1). Hybrid models use attention-layer count, not every layer. If geometry is missing, cache is omitted and that is stated.

1 GiB is added for CUDA/runtime. Apple unified memory uses RAM as the pool. CPU-only is a memory fit, not a speed claim. DDR generation is ignored for fit — it changes bandwidth, not whether the weights sit in memory.

RAM type (DDR4/DDR5/LPDDR) is not a field. It does not change whether a model fits.

Self-host

Questions this page answers

What LLM can I run on a MI300X?

14 of 25 profiled open-weight models fit 192 GiB (vram) at F16, 8192 context. Best rated that fits: Gemma 4 31B.

How much memory does a MI300X have?

192 GiB vendor-published framebuffer (https://www.amd.com/en/products/accelerators/instinct/mi300.html), not free-after-driver.

Does NVIDIA pick the winner?

No. Ranking is arithmetic on sourced parameter counts. Quality leads. Unrated models that fit are listed separately, never scored zero.

Evidence & Ask
Markdown for LLMs