Nemotron 3 Nano Omni 30B A3B

NVIDIA · nvidia/nemotron-3-nano-omni-30b-a3b-reasoning-bf16

Provider	NVIDIA
Family	nemotron
Type	vlm
Status	active
Released	2026-04-20
Updated	2026-08-24
Parameters	33B
Open weights	yes

What it fits on

Ordered by predicted speed. Bandwidth sets decode rate; memory decides whether it runs at all.

Every figure here is computed, not measured. Fit is weights at each quantisation against device memory, with a 25% allowance for the KV cache, activations and the OS. Decode rate is the memory-bandwidth roofline at 70% efficiency. Nobody has run this model on these devices. This is a mixture-of-experts model and no card carries active-parameter counts, so these predictions use total parameters and understate the real speed.

Device	Memory	Bandwidth	Best quality	Weights	tok/s	Fastest
Cerebras WSE-3	44.0 GB	21000000.0 GB/s	q6	24.76 GB	~593658.2	~890487.3 q4
NVIDIA GB200 Grace Blackwell Superchip	372.0 GB	16000.0 GB/s	bf16	66.03 GB	~169.6	~678.5 q4
AMD Instinct MI355X	288.0 GB	8000.0 GB/s	bf16	66.03 GB	~84.8	~339.2 q4
NVIDIA B200	180.0 GB	7700.0 GB/s	bf16	66.03 GB	~81.6	~326.5 q4
Google TPU7x (Ironwood)	192.0 GB	7380.0 GB/s	bf16	66.03 GB	~78.2	~312.9 q4
AMD Instinct MI325X	256.0 GB	6000.0 GB/s	bf16	66.03 GB	~63.6	~254.4 q4
AMD Instinct MI300X	192.0 GB	5300.0 GB/s	bf16	66.03 GB	~56.2	~224.7 q4
NVIDIA H200 SXM	141.0 GB	4800.0 GB/s	bf16	66.03 GB	~50.9	~203.5 q4
NVIDIA H100 NVL	94.0 GB	3900.0 GB/s	bf16	66.03 GB	~41.3	~165.4 q4
NVIDIA H100 SXM	80.0 GB	3350.0 GB/s	fp8	33.02 GB	~71.0	~142.1 q4
AMD Instinct MI250X	128.0 GB	3200.0 GB/s	bf16	66.03 GB	~33.9	~135.7 q4
Google TPU v5p	95.0 GB	2765.0 GB/s	bf16	66.03 GB	~29.3	~117.2 q4
NVIDIA A100 80GB SXM	80.0 GB	2039.0 GB/s	fp8	33.02 GB	~43.2	~86.5 q4
NVIDIA H100 PCIe	80.0 GB	2000.0 GB/s	fp8	33.02 GB	~42.4	~84.8 q4
NVIDIA GeForce RTX 5090	32.0 GB	1792.0 GB/s	q5	20.63 GB	~60.8	~76.0 q4
Google TPU v6e (Trillium)	32.0 GB	1638.0 GB/s	q5	20.63 GB	~55.6	~69.5 q4
AMD Instinct MI210	64.0 GB	1600.0 GB/s	fp8	33.02 GB	~33.9	~67.8 q4
NVIDIA A100 40GB SXM	40.0 GB	1555.0 GB/s	q6	24.76 GB	~44.0	~65.9 q4
Google TPU v4	32.0 GB	1200.0 GB/s	q5	20.63 GB	~40.7	~50.9 q4
NVIDIA GeForce RTX 4090	24.0 GB	1008.0 GB/s	q4	16.51 GB	~42.7	~42.7 q4
AMD Radeon RX 7900 XTX	24.0 GB	960.0 GB/s	q4	16.51 GB	~40.7	~40.7 q4
NVIDIA GeForce RTX 3090	24.0 GB	936.0 GB/s	q4	16.51 GB	~39.7	~39.7 q4
NVIDIA L40S	48.0 GB	864.0 GB/s	fp8	33.02 GB	~18.3	~36.6 q4
Apple M3 Ultra	512.0 GB	819.0 GB/s	bf16	66.03 GB	~8.7	~34.7 q4
Apple M2 Ultra	192.0 GB	800.0 GB/s	bf16	66.03 GB	~8.5	~33.9 q4
Apple M4 Max	128.0 GB	546.0 GB/s	bf16	66.03 GB	~5.8	~23.2 q4
Apple M1 Max	64.0 GB	400.0 GB/s	fp8	33.02 GB	~8.5	~17.0 q4
Apple M2 Max	96.0 GB	400.0 GB/s	bf16	66.03 GB	~4.2	~17.0 q4
Apple M3 Max	128.0 GB	400.0 GB/s	bf16	66.03 GB	~4.2	~17.0 q4
NVIDIA L4	24.0 GB	300.0 GB/s	q4	16.51 GB	~12.7	~12.7 q4
Apple M4 Pro	64.0 GB	273.0 GB/s	fp8	33.02 GB	~5.8	~11.6 q4
Apple M4	32.0 GB	120.0 GB/s	q5	20.63 GB	~4.1	~5.1 q4

Where it runs

Taken from the card's availability section.

Platform	Id on that platform
Hugging Face	nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16

Verified benchmark evidence

Each row was checked against its source by a reviewer, and carries the model identifier as actually evaluated — which is not always the same as this card's.

Benchmark	Evaluated as	Score	Evidence date	Source kind
aa_lcr	Nemotron 3 Nano Omni 30B A3B	39.67%	2026-09-10 evaluated	independent evaluator	source
critpt	Nemotron 3 Nano Omni 30B A3B	0.0%	2026-09-10 evaluated	independent evaluator	source
gdpval_aa	Nemotron 3 Nano Omni 30B A3B	0.0%	2026-09-10 evaluated	independent evaluator	source
gpqa_diamond	Nemotron 3 Nano Omni 30B A3B	46.87%	2026-09-10 evaluated	independent evaluator	source

Reported benchmark scores

This card reports no benchmark scores yet.

Benchmark	Catalogue standing	Score

Data

This card as JSON · See it in the graph · Edit on GitHub