Nemotron 3 Nano Omni 30B-A3B vs Qwen3.8-27B

Qwen3.8-27B leads on the Visual Intelligence Index by 8.2 points, and Nemotron 3 Nano Omni 30B-A3B scores higher on 0 of the 7 benchmarks where the two differ.

Model

Compared with

NVIDIANemotron 3 Nano Omni 30B-A3B60.3%Rank #8 AlibabaQwen3.8-27B68.5%Rank #6

At a glance

Measure	Nemotron 3 Nano Omni 30B-A3B	Qwen3.8-27B
Visual Intelligence Index	60.3%	68.5%
Rank	#8 of 9	#6 of 9
Cost per taskMean over the benchmarks, at list price.	—	—
Latency per taskEach benchmark's median, averaged.	14.7s	32.0s
Tokens per task	35,341	25,009
Input / 1M tokens	—	—
Output / 1M tokens	—	—
Context window	262K	256K
Availability	Open weights	Open weights

How cost and latency are measured

Capabilities

Each capability index is a weighted mean of the benchmarks that contribute to the capability.

Capability	Nemotron 3 Nano Omni 30B-A3B	Qwen3.8-27B	Lead
Perception	61.0%	67.8%	−6.8
Spatial Reasoning	62.0%	68.6%	−6.6
Temporal Reasoning	60.2%	68.4%	−8.2
Action / Event Understanding	55.6%	65.2%	−9.6
Causal Reasoning	63.6%	72.3%	−8.7
Information Retrieval	57.1%	67.0%	−9.9

Benchmarks

Benchmark	Nemotron 3 Nano Omni 30B-A3B	Qwen3.8-27B	Lead
Video-MME v2	35.0%	50.5%	−15.6
LVBench	54.2%	64.6%	−10.3
Perception Test	70.3%	72.3%	−2.0
NExT-QA	81.2%	82.2%	−1.0
Q-Bench Video	64.9%	67.9%	−3.0
EgoSchema	65.8%	79.1%	−13.3
UCF101-AD	31.0%	45.5%	−14.5

Usability

Nemotron 3 Nano Omni 30B-A3B

5.0/ 5

Reachable1.0: There is an endpoint you can call without hosting anything.
First-party API — supported
Third-party API — supported
Portable1.0: You can run it yourself, and are not tied to one vendor.
Open weights — supported
Self-hostable — supported
Multimodal input1.0: It takes the footage directly, rather than frames you extracted.
Video ingestion — supported
Audio — supported
Programmable1.0: Output you can parse, and tools it can call on its own.
Structured output — supported
Tool calling — supported
Operable1.0: Usable interactively and in bulk, not only one call at a time.
Streaming — supported
Batch API — supported

Qwen3.8-27B

4.5/ 5

Reachable1.0: There is an endpoint you can call without hosting anything.
First-party API — supported
Third-party API — supported
Portable1.0: You can run it yourself, and are not tied to one vendor.
Open weights — supported
Self-hostable — supported
Multimodal input1.0: It takes the footage directly, rather than frames you extracted.
Video ingestion — supported
Audio — supported
Programmable1.0: Output you can parse, and tools it can call on its own.
Structured output — supported
Tool calling — supported
Operable0.5: Usable interactively and in bulk, not only one call at a time.
Streaming — supported
Batch API — not supported