Can Ryzen AI Max+ 395 128GB run Devstral-Small-2-24B-Instruct-2512?

AMD · 96 GB LPDDR5X · 215 GB/s bandwidth · Mistral · 24.0B · model ctx up to 256K

ROCM

Yes — the Ryzen AI Max+ 395 128GB (96 GB LPDDR5X) can run Devstral-Small-2-24B-Instruct-2512. At Q4 with the default 8K context it needs ≈ 19.0 GB VRAM (weights 13.2 GB + KV cache 4.8 GB + 1 GB runtime overhead), which fits within 96 GB. Decode at Q4/8K: ≈ 5.7 tok/s (estimated, batch 1). FP16 (≈ 53.8 GB) also fits, at ≈ 1.6 tok/s (estimated). Estimates come from memory-bandwidth math; rows tagged measured override estimates. Relative ranking is more reliable than absolute tok/s. Methodology.

Fit & speed by quant and context

Quant	Context	VRAM needed	Fits?	tok/s (decode)
Q4	4K	16.6 GB	Yes	5.7 est.
Q4	8K default	19.0 GB	Yes	5.7 est.
Q4	32K	33.4 GB	Yes	5.7 est.
FP16	4K	51.4 GB	Yes	1.6 est.
FP16	8K default	53.8 GB	Yes	1.6 est.
FP16	32K	68.2 GB	Yes	1.6 est.

Fits = weights + KV(ctx) + 1 GB overhead ≤ 95% of VRAM. Measured anchors are context-agnostic; the fit verdict is recomputed per context. A missing tok/s means the model is far beyond this card (offload-only territory).

VRAM breakdown at 8K context

Quant	Weights	KV cache	Overhead	Total needed	Ryzen AI Max+ 395 128GB VRAM
Q4	13.2 GB	4.8 GB	1 GB	19.0 GB	96 GB
FP16	48.0 GB	4.8 GB	1 GB	53.8 GB	96 GB

Other GPUs that run Devstral-Small-2-24B-Instruct-2512

• B200 192GB SXM — ≈212.1 tok/s Q4 est.
• MI300X 192GB — ≈140.5 tok/s Q4 est.
• H200 141GB SXM5 — ≈127.3 tok/s Q4 est.
• H100 80GB SXM5 — ≈88.8 tok/s Q4 est.
• Gaudi 3 128GB — ≈86.9 tok/s Q4 est.
• A100 80GB SXM4 — ≈54.1 tok/s Q4 est.
• RTX 5090 32GB — ≈47.5 tok/s Q4 est.
• RTX 4090 24GB — ≈26.7 tok/s Q4 est.

All GPUs for Devstral-Small-2-24B-Instruct-2512 →

Other models for the Ryzen AI Max+ 395 128GB

• MiniCPM5-1B — ≈126.7 tok/s Q4 est.
• Qwen3.5-2B — ≈68.4 tok/s Q4 est.
• DeepSeek-R1-0528-Qwen3-8B-layer-mix-bpw-3.8-mlx — ≈59.5 tok/s Q4 est.
• Qwen3.6-35B-A3B — ≈45.6 tok/s Q4 est.
• Qwen3.5-35B-A3B — ≈45.6 tok/s Q4 est.
• GLM-4.7-Flash — ≈45.6 tok/s Q4 est.
• Qwen3-30B-A3B-Instruct-2507 — ≈45.6 tok/s Q4 est.
• Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 — ≈45.6 tok/s Q4 est.

All models on the Ryzen AI Max+ 395 128GB →