Downloads · 30 days
61
35% of all-time downloads
sorbet-25m-gguf
CodeSoft/sorbet-25m-gguf
sorbet-25m-gguf is a text generation model from CodeSoft. Use it when you need the model to write or continue text. The card lists the license as apache-2.0.
This repository contains GGUF quantizations of CodeSoft/sorbet-25m for use with llama.cpp.
All-time downloads
175
Public
Repo size
154 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.gguf154 MB · 100%
At a glance
- Task
- Text Generation
- License
- apache-2.0
- Access
- Public
- Created
- Aug 22, 2026
- Updated
- Aug 22, 2026
- SHA
- dc7702a9
Try a prompt
Base models
Model card
From the Hugging Face model README
Sorbet 25M GGUF
This repository contains GGUF quantizations of CodeSoft/sorbet-25m for use with llama.cpp.
Quantizations
All XL quants are dynamic mixes calibrated with an importance matrix (imatrix from 1M tokens of the pretraining mix).
| File | Quantization | Size |
|---|---|---|
| Sorbet-25M-BF16.gguf | BF16 | 50.7MB |
| Sorbet-25M-F16.gguf | F16 | 50.7MB |
| Sorbet-25M-Q8_K_XL.gguf | Q8_0 weights (imatrix) + F16 tied embd/output + F32 norms | 30.0MB |
| Sorbet-25M-Q4_K_XL.gguf | Q4_0 gate/up (imatrix) + Q5_0 attn_output + Q8_0 attn_q/k/v + Q6_K ffn_down + F16 tied embd/output + F32 norms | 22.4MB |
Perplexity on the 1M-token calibration split: F16 = 45.13, BF16 = 45.13, Q8_K_XL = 45.15 (+0.02), Q4_K_XL = 46.22 (+1.09).
- Task
- Text Generation
- License
- apache-2.0
- Languages
- en
- Created
- Aug 22, 2026
- Updated
- Aug 22, 2026