NVIDIA Nemotron Labs Dory 1.0 vs Visual Jev (Qwen3-VL 8B Backbone)
A direct, empirical evaluation of NVIDIA Nemotron Labs Dory 1.0 (NVIDIA) against Visual Jev (Qwen3-VL 8B Backbone) (Independent AI Systems Lab (Guanxu Yu & Yuhang Yao)).
NVIDIA Nemotron Labs Dory 1.0
Developed by NVIDIA · 4 Billion Parameters (Looped Transformer) parameters
Ultra-low VRAM memory residency (4.4 GB FP8), 256k long-context needle retrieval, and recurrent layer depth execution.
Visual Jev (Qwen3-VL 8B Backbone)
Developed by Independent AI Systems Lab (Guanxu Yu & Yuhang Yao) · 8.2 Billion Vision-Language Parameters parameters
Sub-20ms multi-query image verification, automated rejection sampling in generative diffusion pipelines (FLUX / SD3), robotics perception routing, and high-frequency visual software inspection.
| Metric | NVIDIA Nemotron Labs Dory 1.0 | Visual Jev (Qwen3-VL 8B Backbone) |
|---|---|---|
| SWE-bench Verified | 66.4% | 62.1% |
| MMLU-Pro | 82.8% | 86.4% |
| MATH-500 | 85.6% | 89.2% |
| Context Window | 256,000 tokens | 32,768 tokens (Visual prefix + batched question suffixes) |
| Input Pricing (per 1M) | $0.15 | $0.05 |
| Output Pricing (per 1M) | $0.45 | $0.10 |
Architectural & Practical Deployment Analysis
Evaluating NVIDIA Nemotron Labs Dory 1.0 (NVIDIA) alongside Visual Jev (Qwen3-VL 8B Backbone) (Independent AI Systems Lab (Guanxu Yu & Yuhang Yao)) highlights key technical trade-offs across neural parameter scale and compute efficiency. NVIDIA Nemotron Labs Dory 1.0 executes across 4 Billion Parameters (Looped Transformer) total parameters with 4.1B Active Recurrent Block (Equivalent to 16B Depth) active per token, whereas Visual Jev (Qwen3-VL 8B Backbone) operates on 8.2 Billion Vision-Language Parameters parameters with 8.2B (Pinned Shared Prefix KV Cache) active per token.
On SWE-bench Verified software engineering tasks, NVIDIA Nemotron Labs Dory 1.0 achieves 66.4% autonomous resolution rate, compared to 62.1% for Visual Jev (Qwen3-VL 8B Backbone). On complex reasoning benchmarks, NVIDIA Nemotron Labs Dory 1.0 posts 82.8% on MMLU-Pro and 85.6% on MATH-500, against 86.4% and 89.2% respectively for Visual Jev (Qwen3-VL 8B Backbone).
From an operational economics perspective, processing 10,000,000 mixed tokens (7M input / 3M output) costs approximately $2.40 on NVIDIA Nemotron Labs Dory 1.0 versus $0.65 on Visual Jev (Qwen3-VL 8B Backbone).
Recommendation: Deploy NVIDIA Nemotron Labs Dory 1.0 when your infrastructure prioritizes ultra-low vram memory residency (4.4 gb fp8), 256k long-context needle retrieval, and recurrent layer depth execution.. Deploy Visual Jev (Qwen3-VL 8B Backbone) when workload requirements favor sub-20ms multi-query image verification, automated rejection sampling in generative diffusion pipelines (flux / sd3), robotics perception routing, and high-frequency visual software inspection..