V100 16GB PCIe — local LLM fit & speed

NVIDIA · 16 GB HBM2 · 900 GB/s · 250 W · MSRP $8,000

CUDA

The V100 16GB PCIe is a NVIDIA accelerator with 16 GB HBM2 and about 900 GB/s peak memory bandwidth. AI Hashrate estimates local LLM decode (batch 1) at 8K context: fit counts weights + KV cache + 1 GB runtime overhead — not weights alone. At Q4, about 24 curated models fit fully on this card; at FP16, about 8 fit. Top Q4 speeds: MiniCPM5-1B ≈ 530.3 tok/s (estimated, fits); Qwen3.5-2B ≈ 286.4 tok/s (estimated, fits); DeepSeek-R1-0528-Qwen3-8B-layer-mix-bpw-3.8-mlx ≈ 249.0 tok/s (estimated, fits); Llama-3.2-3B-Instruct ≈ 179.0 tok/s (estimated, fits); gpt-oss-20b ≈ 159.1 tok/s (estimated, fits). Relative ranking is more reliable than absolute tok/s. See methodology for the bandwidth formula and measured-anchor policy. Methodology.

Table: decode speed estimates (Q4 / FP16). Measured rows override estimates. MSRP is list price, not live retail. Context default 8K.

Model	Params (B)	Quant	tok/s	Fits?
MiniCPM5-1B	1.08	Q4	530.3 est.	Yes
Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16	30.0	Q4	453.0 est.	No
Qwen3-30B-A3B-Instruct-2507	30.5	Q4	412.0 est.	No
GLM-4.7-Flash	31.2	Q4	395.5 est.	No
Qwen3.5-35B-A3B	34.7	Q4	342.5 est.	No
qwen35b-a3b-fable-sft-abliterated	34.7	Q4	325.2 est.	No
Ornith-1.0-35B	35.0	Q4	320.2 est.	No
Qwen3.6-35B-A3B	36.0	Q4	319.3 est.	No
Qwen3.5-2B	2.0	Q4	286.4 est.	Yes
DeepSeek-R1-0528-Qwen3-8B-layer-mix-bpw-3.8-mlx	2.3	Q4	249.0 est.	Yes
Llama-3.2-3B-Instruct	3.2	Q4	179.0 est.	Yes
gpt-oss-20b	21.5	Q4	159.1 est.	Yes
gemma-4-26B-A4B-it	25.2	Q4	150.7 est.	Yes
MiniCPM5-1B	1.08	FP16	145.8 est.	Yes
Qwen3-4B-Instruct-2507	4.0	Q4	143.2 est.	Yes
NVIDIA-Nemotron-3-Nano-4B-BF16	4.0	Q4	143.2 est.	Yes
Qwen3-4B-Thinking-2507	4.0	Q4	143.2 est.	Yes
gemma-4-E2B-it	5.1	Q4	112.3 est.	Yes
Mistral-7B-Instruct-v0.3	7.2	Q4	79.5 est.	Yes
Qwen3.5-2B	2.0	FP16	78.8 est.	Yes
Qwen2.5-7B-Instruct	7.6	Q4	75.4 est.	Yes
Llama-3.1-8B-Instruct	8.0	Q4	71.6 est.	Yes
gemma-4-E4B-it	8.0	Q4	71.6 est.	Yes
Qwen3-8B	8.2	Q4	69.8 est.	Yes
DeepSeek-R1-0528-Qwen3-8B	8.2	Q4	69.8 est.	Yes
Qwen3-Coder-Next	79.7	Q4	69.0 est.	No
DeepSeek-R1-0528-Qwen3-8B-layer-mix-bpw-3.8-mlx	2.3	FP16	68.5 est.	Yes
Qwen3-Next-80B-A3B-Instruct	81.3	Q4	66.4 est.	No
Qwen3.6-35B-A3B-NVFP4	21.6	Q4	65.4 est.	No
Qwen3.5-9B	8.95	Q4	64.0 est.	Yes
Ornith-1.0-9B	9.0	Q4	63.6 est.	Yes
Qwen3.5-27B	27.0	Q4	56.6 est.	No
Mistral-7B-Instruct-v0.3	7.2	FP16	56.2 est.	No
Qwen3.6-27B	27.8	Q4	52.1 est.	No
Llama-3.2-3B-Instruct	3.2	FP16	49.2 est.	Yes
Devstral-Small-2-24B-Instruct-2512	24.0	Q4	48.2 est.	No
Qwen2.5-7B-Instruct	7.6	FP16	48.1 est.	No
gemma-4-12b-it	11.95	Q4	47.9 est.	Yes
OLMo-2-1124-13B-Instruct	13.7	Q4	41.8 est.	Yes
Llama-3.1-8B-Instruct	8.0	FP16	41.5 est.	No
gemma-4-E4B-it	8.0	FP16	41.5 est.	No
Qwen3-4B-Instruct-2507	4.0	FP16	39.4 est.	Yes
NVIDIA-Nemotron-3-Nano-4B-BF16	4.0	FP16	39.4 est.	Yes
Qwen3-4B-Thinking-2507	4.0	FP16	39.4 est.	Yes
Qwen3-14B	14.8	Q4	38.7 est.	Yes
Qwen2.5-14B-Instruct	14.8	Q4	38.7 est.	Yes
DeepSeek-R1-Distill-Qwen-14B	14.8	Q4	38.7 est.	Yes
Qwen3-8B	8.2	FP16	38.6 est.	No
DeepSeek-R1-0528-Qwen3-8B	8.2	FP16	38.6 est.	No
gemma-4-26B-A4B-it-AWQ-4bit	26.6	Q4	35.8 est.	No
Qwen3.5-9B	8.95	FP16	35.1 est.	No
gemma-3-27b-it	27.4	Q4	32.8 est.	No
gemma-4-E2B-it	5.1	FP16	30.9 est.	Yes
Ornith-1.0-9B	9.0	FP16	29.5 est.	No
Qwen3.5-122B-A10B-Heretic-v2-MLX-mixed-3.8bit	19.8	Q4	28.9 est.	Yes
gemma-4-31B-it	30.7	Q4	23.6 est.	No
gemma-4-31B-it-FP8-block	31.3	Q4	22.3 est.	No
Qwen2.5-32B-Instruct	32.8	Q4	19.4 est.	No
Qwen2.5-Coder-32B-Instruct	32.8	Q4	19.4 est.	No
DeepSeek-R1-Distill-Qwen-32B	32.8	Q4	19.4 est.	No
gpt-oss-20b	21.5	FP16	16.0 est.	No
Qwen3.6-35B-A3B-FP8	36.0	Q4	14.8 est.	No
gemma-4-12b-it	11.95	FP16	12.9 est.	No
gemma-4-26B-A4B-it	25.2	FP16	11.1 est.	No
Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16	30.0	FP16	10.3 est.	No
Qwen3-30B-A3B-Instruct-2507	30.5	FP16	9.8 est.	No
GLM-4.7-Flash	31.2	FP16	9.3 est.	No
OLMo-2-1124-13B-Instruct	13.7	FP16	8.6 est.	No
Qwen3-14B	14.8	FP16	6.9 est.	No
Qwen2.5-14B-Instruct	14.8	FP16	6.9 est.	No
DeepSeek-R1-Distill-Qwen-14B	14.8	FP16	6.9 est.	No
Qwen3.5-122B-A10B-Heretic-v2-MLX-mixed-3.8bit	19.8	FP16	2.9 est.	No
Qwen3.6-35B-A3B-NVFP4	21.6	FP16	2.3 est.	No
Llama-3.1-70B-Instruct	70.6	Q4	2.1 est.	No
Llama-3.3-70B-Instruct	70.6	Q4	2.1 est.	No
Qwen2.5-72B-Instruct	72.7	Q4	1.9 est.	No
Devstral-Small-2-24B-Instruct-2512	24.0	FP16	1.7 est.	No
Qwen3.5-27B	27.0	FP16	1.4 est.	No
Qwen3.6-27B	27.8	FP16	1.3 est.	No
gemma-4-26B-A4B-it-AWQ-4bit	26.6	FP16	1.2 est.	No
gemma-3-27b-it	27.4	FP16	1.1 est.	No

Fit checks

Qwen3-8B — fits · ≈69.8 tok/s Q4 est.
gemma-4-26B-A4B-it — fits · ≈150.7 tok/s Q4 est.
Qwen2.5-7B-Instruct — fits · ≈75.4 tok/s Q4 est.
Qwen3.5-9B — fits · ≈64.0 tok/s Q4 est.
Qwen3.6-35B-A3B-NVFP4 — does not fit · ≈65.4 tok/s Q4 est.
Llama-3.1-8B-Instruct — fits · ≈71.6 tok/s Q4 est.
Qwen3.6-35B-A3B-FP8 — does not fit · ≈14.8 tok/s Q4 est.
gpt-oss-20b — fits · ≈159.1 tok/s Q4 est.
Qwen3.6-35B-A3B — does not fit · ≈319.3 tok/s Q4 est.
Qwen3.5-35B-A3B — does not fit · ≈342.5 tok/s Q4 est.
gemma-4-E2B-it — fits · ≈112.3 tok/s Q4 est.

All 73 model fit checks — see the linked Fits? column in the table above