Spaces:
Configuration error
Configuration error
File size: 1,476 Bytes
ab51d1c 0d314c9 27d655d 1e2fc52 27d655d 1e2fc52 ab51d1c 0cbb258 0d314c9 0cbb258 1e2fc52 0cbb258 1e2fc52 62a6e5c 0cbb258 27d655d 1e2fc52 7eb59ff 1e2fc52 7c6d065 1e2fc52 | 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 | ---
title: ThingAI
type: org
tags:
- slm
- llm
- pytorch
- bash
- shell
- code
- tokenizer
- rust
---
# ThingsAI
Building efficient, specialist Small Language Models that run on consumer hardware. Zero telemetry. Open weights. Everything from tokenizer to training script is public.
## Models
* **Dwarf-15M**
A 15.54M parameter shell/bash specialist. 12 layers, d_model=320, GQA 5Q/1KV, SwiGLU, RMSNorm, RoPE. Custom 8202-token vocabulary via DwarfGoToken. 1390:1 token-to-parameter ratio across 11 datasets spanning raw shell, Python, C, instruction pairs, and English web text. Target use case: CLI tool that translates natural language into bash commands with user review before execution.
* **Quark-270M**
Our largest model. 252M effective parameters, 32 layers, d_model=768, GQA 12Q/4KV, 65K bilingual vocabulary (Italian + English). Trained on curated multilingual data. Available as Base and Instruct variants.
* **Quark-135M**
Bilingual (Italian + English) general-purpose model. 135M parameters, 30 layers, 9 attention heads (3 KV, GQA), SwiGLU, RMSNorm, RoPE θ=10k. Trained on 15B+ tokens. Published benchmarks: HellaSwag 31.37%, ARC-Easy 41.46%, PIQA 61.26%.
## Links
* Models and tokenizers: [HuggingFace](https://huggingface.co/ThingAI)
* Script & Tool: [GitHub](https://github.com/overcastlab)
* Website: [things-ai.org](https://things-ai.org)
* GoToken: [crates.io](https://crates.io/crates/gotoken) · [PyPI](https://pypi.org/project/gotoken/) |