Can Ryzen AI Max+ 395 64GB run Mistral-7B-Instruct-v0.3?

AMD · 48 GB LPDDR5X · 215 GB/s bandwidth · Mistral · 7.2B · model ctx up to 32K

ROCM

Yes — the Ryzen AI Max+ 395 64GB (48 GB LPDDR5X) can run Mistral-7B-Instruct-v0.3. At Q4 with the default 8K context it needs ≈ 6.4 GB VRAM (weights 3.96 GB + KV cache 1.44 GB + 1 GB runtime overhead), which fits within 48 GB. Decode at Q4/8K: ≈ 19.0 tok/s (estimated, batch 1). FP16 (≈ 16.84 GB) also fits, at ≈ 5.2 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	5.68 GB	Yes	19.0 est.
Q4	8K default	6.4 GB	Yes	19.0 est.
Q4	32K	10.72 GB	Yes	19.0 est.
FP16	4K	16.12 GB	Yes	5.2 est.
FP16	8K default	16.84 GB	Yes	5.2 est.
FP16	32K	21.16 GB	Yes	5.2 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 64GB VRAM
Q4	3.96 GB	1.44 GB	1 GB	6.4 GB	48 GB
FP16	14.4 GB	1.44 GB	1 GB	16.84 GB	48 GB

Other GPUs that run Mistral-7B-Instruct-v0.3

• B200 192GB SXM — ≈707.1 tok/s Q4 est.
• MI300X 192GB — ≈468.4 tok/s Q4 est.
• H200 141GB SXM5 — ≈424.2 tok/s Q4 est.
• H100 80GB SXM5 — ≈296.1 tok/s Q4 est.
• Gaudi 3 128GB — ≈289.5 tok/s Q4 est.
• RTX 5090 32GB — ≈255.9 tok/s Q4 measured
• RTX 4090 24GB — ≈183.3 tok/s Q4 measured
• A100 80GB SXM4 — ≈180.2 tok/s Q4 est.

All GPUs for Mistral-7B-Instruct-v0.3 →

Other models for the Ryzen AI Max+ 395 64GB

• 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 64GB →