Can B200 192GB SXM run Qwen3.5-122B-A10B-Heretic-v2-MLX-mixed-3.8bit?

NVIDIA · 192 GB HBM3e · 8000 GB/s bandwidth · TheCluster · 19.8B (active 19.8B) · MoE · model ctx up to NoneK

CUDA

Yes — the B200 192GB SXM (192 GB HBM3e) can run Qwen3.5-122B-A10B-Heretic-v2-MLX-mixed-3.8bit. At Q4 with the default 8K context it needs ≈ 15.85 GB VRAM (weights 10.89 GB + KV cache 3.96 GB + 1 GB runtime overhead), which fits within 192 GB. Decode at Q4/8K: ≈ 257.1 tok/s (estimated, batch 1). FP16 (≈ 44.56 GB) also fits, at ≈ 70.7 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	13.87 GB	Yes	257.1 ± est.
Q4	8K default	15.85 GB	Yes	257.1 ± est.
Q4	32K	27.73 GB	Yes	257.1 ± est.
FP16	4K	42.58 GB	Yes	70.7 ± est.
FP16	8K default	44.56 GB	Yes	70.7 ± est.
FP16	32K	56.44 GB	Yes	70.7 ± 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	B200 192GB SXM VRAM
Q4	10.89 GB	3.96 GB	1 GB	15.85 GB	192 GB
FP16	39.6 GB	3.96 GB	1 GB	44.56 GB	192 GB

Other GPUs that run Qwen3.5-122B-A10B-Heretic-v2-MLX-mixed-3.8bit

• MI300X 192GB — ≈170.3 tok/s Q4 est.
• H200 141GB SXM5 — ≈154.3 tok/s Q4 est.
• H100 80GB SXM5 — ≈107.7 tok/s Q4 est.
• Gaudi 3 128GB — ≈105.3 tok/s Q4 est.
• A100 80GB SXM4 — ≈65.5 tok/s Q4 est.
• RTX 5090 32GB — ≈57.6 tok/s Q4 est.
• RTX 4090 24GB — ≈32.4 tok/s Q4 est.
• RTX 4090D 24GB (China) — ≈32.4 tok/s Q4 est.

All GPUs for Qwen3.5-122B-A10B-Heretic-v2-MLX-mixed-3.8bit →

Other models for the B200 192GB SXM

• MiniCPM5-1B — ≈4713.8 tok/s Q4 est.
• Qwen3.5-2B — ≈2545.5 tok/s Q4 est.
• DeepSeek-R1-0528-Qwen3-8B-layer-mix-bpw-3.8-mlx — ≈2213.4 tok/s Q4 est.
• Qwen3.6-35B-A3B — ≈1697.0 tok/s Q4 est.
• Qwen3.5-35B-A3B — ≈1697.0 tok/s Q4 est.
• GLM-4.7-Flash — ≈1697.0 tok/s Q4 est.
• Qwen3-30B-A3B-Instruct-2507 — ≈1697.0 tok/s Q4 est.
• Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 — ≈1697.0 tok/s Q4 est.

All models on the B200 192GB SXM →