#4545th of 90 in Open models
eagle
Vision-language model family from NVIDIA: Eagle, Eagle 2.5, LocateAnything
Live demo ↗
(opens in a new tab)Website ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 3.9k
- License
- Apache-2.0
- Last commit
- Nov 2026
- Language
- Python
936 on 25 Apr 20263,855 on 14 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Code and checkpoints for NVIDIA's Eagle vision-language models: Eagle, Eagle 2, Eagle 2.5 (long-context image and video understanding), LocateAnything (grounding and detection), and VFlash (visual planning for robotics). Weights are on Hugging Face, built on Qwen2.5 or Llama-3 backbones with SigLIP, ConvNeXt or MoonViT encoders. The code is Apache-2.0, while the model weights use a separate NVIDIA license.
Who it is for: Researchers and engineers working on vision-language models and robotics
Strengths
- Several model generations in one repo, from Eagle2-1B to Eagle2.5-8B
- Eagle2.5-8B lists a 128K max length for long video and multi-image input
- LocateAnything batch inference runs on FlashAttention, including A100 and RTX 4090
- LoRA fine-tuning script provided for LocateAnything visual prompts
Weaknesses
- Model weights use the NVIDIA license, not Apache-2.0, which may limit use
- Research repo with no Docker files and no tagged release
- Setup and docs are split across separate subfolders per model
- Top-level README gives no RAM or VRAM requirements
What it needs
- Needs Hugging Face model hub
- Models: Eagle, Eagle 2, Eagle 2.5, LocateAnything-3B, VFlash
Also in Open models
See all 90| Rank | Project | Score |
|---|---|---|
| 1 | ruviewWiFi CSI sensing on ESP32 for presence, breathing, heart rate and pose | 94 out of 100 |
| 2 | layaSingle-pass multilingual classifier returning typed decisions with calibrated odds | 88 out of 100 |
| 3 | rf-detrReal-time transformer models for object detection, segmentation and keypoints | 82 out of 100 |
| #4 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 80 out of 100 |
| #5 | open_clipPyTorch library for training and running CLIP-style image-text models | 79 out of 100 |