384M English model trained from scratch on ~17B tokens. Thinks step by step, holds a conversation, reads ~16K tokens. Apache-2.0.
AI & ML interests
AI research from first principles.
Recent Activity
View all activity
Organization Card
Udgam Labs
AI research from first principles.
| Family | What it does | Start here |
|---|---|---|
| Gargi | English chat model: thinks step by step, holds a conversation, reads ~16K tokens; 384M parameters, ~1,000× less training data than Qwen2.5-0.5B | Gargi · GGUF · MLX · try it |
| Lilavati | Maths word problems: writes a short Python program and runs it; 95M parameters | Lilavati · Checker · try it |
| Chitra | Text-to-image: 212M parameters, 256×256 images | Chitra-Tiny · try it |
| Udgam-384M | English on-device tool calling: app actions, document → JSON, shell commands | Udgam-384M-Chat · GGUF · MLX |
| Early experiments | Our first releases, kept public for transparency (one is a Qwen3-0.6B adapter) | udgam-english-125m · hinglish-commands-0.6b |
Each family has its own collection on this page.
models 16
udgamlabs/udgam-english-125m
Text Generation • Updated • 1
udgamlabs/hinglish-commands-0.6b
Text Generation • Updated • 1
udgamlabs/Chitra-Tiny
Text-to-Image • 0.2B • Updated • 20 • 3
udgamlabs/Lilavati-Checker
Text Generation • 94.7M • Updated • 300
udgamlabs/Lilavati
Text Generation • 94.7M • Updated • 618
udgamlabs/Udgam-384M-Chat-MLX-4bit
Text Generation • 0.4B • Updated • 81
udgamlabs/Udgam-384M-Chat-MLX
Text Generation • 0.4B • Updated • 1.06k • 1
udgamlabs/Udgam-384M-Chat-GGUF
Text Generation • 0.4B • Updated • 371 • 1
udgamlabs/Udgam-384M-Chat
Text Generation • 0.4B • Updated • 1.01k
udgamlabs/Udgam-384M-Base-4K
Text Generation • 0.4B • Updated • 1.08k
datasets 0
None public yet