Cactus-Compute/gemma-4-e2b-it-hybrid
Text Generation • 5B • Updated • 146 • 2
Cactus Hybrid
A small, on-device model is fast and private, but sometimes wrong. We post-train models to know when they are wrong, run on any framework.
Note Transformers checkpoint. generate(..., return_confidence=True) returns the probe confidence. pip install 'transformers>=5.5.4,<5.6' torch
Cactus-Compute/gemma-4-e2b-it-hybrid-GGUF
5B • Updated • 530 • 2Note GGUF quants for the patched llama.cpp server (patch in the GitHub repo until upstreamed): responses carry a top-level confidence field.
Cactus-Compute/gemma-4-e2b-it-hybrid-mlx
Text Generation • 0.7B • Updated • 167 • 2Note MLX weights. model.last_confidence holds the probe confidence after generate(). pip install mlx-lm
Cactus-Compute/gemma-4-E2B-it
Image-Text-to-Text • Updated • 239 • 7Note Cactus engine bundle. cactus_complete() returns confidence alongside the response. pip install cactus-compute