NVLM 1.0
72B frontier-class multimodal model that improved its own text backbone
About
NVLM 1.0 is NVIDIA's multimodal family from late 2024, led by the 72B NVLM-D model built on Qwen2-72B with an InternViT-6B vision encoder. It was notable for two things: rivaling GPT-4o on vision-language tasks, and actually improving text-only performance after multimodal training (up 4.3 points on average across text benchmarks), where most VLMs regress. NVIDIA reports 58.7 on MMMU, 65.2 on MathVista, and 852 on OCRBench. Weights are public on Hugging Face, but under CC BY-NC 4.0, which bars commercial use, so it is a research artifact rather than a production model for most companies.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
MMMU 58.7, MathVista 65.2, OCRBench 852 (NVIDIA-reported)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: NVLM-D-72B paper Sept 2024, weights on Hugging Face verified via NVIDIA coverage and HF model card; MMMU 58.7, MathVista 65.2, OCRBench 852 are NVIDIA-reported; labeled as such; CC BY-NC 4.0 license bars commercial use; caveat captured prominently.