Compare GPU memory, context windows, and deployment specs for the latest open-weight models.
Recent releases, picked by BenchGrid.
A mid-sized vision-language model for reasoning, coding, and agent workloads.
A compact Qwen-based MiMo distillation for a smaller deployment footprint.
Unified text, image, and audio understanding in a compact Gemma 4 model.
A sparse model with extra embedding memory for long-context agent workloads.
MiMo's multimodal Flash model with sparse experts and a 1M context window.
A multimodal GLM model combining sparse experts with hybrid attention.
Specifications from official model cards. Memory figures are calculated estimates, not measured VRAM. How to read the data
Start with a deployment question.
How much space do the weights take?
Which models advertise room for longer inputs?
Where should a smaller deployment experiment start?
