File size: 1,487 Bytes
9552a47
 
c3c6533
 
 
9552a47
 
 
 
c3c6533
f844d50
c3c6533
 
 
 
 
01287c2
c3c6533
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
---
title: README
emoji: ⬛
colorFrom: gray
colorTo: gray
sdk: static
pinned: false
---

<p align="center">
  <img src="https://huggingface.co/spaces/SectileLabs/README/resolve/main/sectile-banner.png" alt="Sectile Research Laboratories. We build the model, and everything it runs on." width="100%">
</p>

**Sectile Research Laboratories** is rebuilding the AI stack from the representation up. We release consumer-focused versions of the systems we run on to the open-source and local-hosting community.

<p align="center">
  <img src="https://huggingface.co/spaces/SectileLabs/README/resolve/main/products.png" alt="GInfer: open-source inference engine for fast, high-concurrency, long-context work on NVIDIA GPUs, Apache 2.0. GChat: the desktop app for GInfer on Windows and Linux." width="100%">
</p>

## Models

The [GInfer Models](https://huggingface.co/collections/SectileLabs/ginfer-models-6aa0682395e628be98a798ef) collection holds our model packages for GInfer and GChat. Each package's card names its base model and license.

| Format | What it is | GPUs |
|---|---|---|
| INT | 4-bit groupwise integer | Ampere, Ada, Blackwell |
| NVFP4 | 4-bit NVIDIA FP4 | Blackwell (RTX 50 series, RTX PRO) |
| SMOL | 2-bit, mixed with 4-bit where a model needs it | Ampere, Ada, Blackwell |

The packages run only in GInfer and GChat.

## Learn more

- [sectilelabs.ai](https://sectilelabs.ai)
- [Case study: building GInfer with Quaestor](https://sectilelabs.ai/case-study-ginfer.html)