About TeichAI
A small lab with a simple loop: ask the best models, teach the open ones, publish everything.
Frontier models from Anthropic, OpenAI, Google, DeepSeek and others are extraordinary, but they live behind APIs. We think their reasoning style should be something you can run on your own hardware. So we capture it, distill it into open-weight bases, and give the whole chain away.
- Model releases
- 66
- Datasets
- 54
- All-time downloads
- 2.2M
- HF followers
- 1.5K
The people
Four of us, a lot of GPU hours.
Everything here is a side project funded out of pocket. Say hi on Hugging Face.
How we work
One pipeline, end to end.
- 01
Curate prompts
Coding, math, science and open-ended tasks. For agent traces we point real repositories at the teacher and let it work.
prompts.jsonl
- 02
Generate traces
Frontier models answer with maximum reasoning effort. Thinking, tool calls and results are captured with teich and published as a dataset.
teich generate
- 03
Fine-tune an open base
Qwen, Gemma, GLM, gpt-oss and friends are trained with Unsloth. Response-only masking keeps the loss on what the teacher actually said.
mask_data()
- 04
Quantize and ship
Safetensors plus imatrix GGUF from Q2 to Q8, vision projectors when the base sees images, and an Ollama Modelfile. Apache-2.0.
llama-server -hf
What we stand for
Principles we try not to break.
Open weights, open data
Every model ships with the dataset that made it. If you can download the distill, you can download what it learned from and retrain it yourself.
Open tooling
The generation, extraction and masking pipeline is published as teich on PyPI. Our recipes are reproducible, not folklore.
Honest model cards
Base model, dataset, training setup and known limitations are in every card. We would rather be boring than overclaim.
Community-directed
Distill requests come in through Discord. A lot of what we ship started as someone asking for it.
Get involved
Join the community or fund the next run.
Discord
Distill requests, release announcements, training chatter. The fastest way to reach us.
Join the serverHugging Face
1.5K followers get every drop first. Weights, quants and datasets live here.
Follow TeichAISupport the lab
Every coffee turns into API credits and GPU hours.
We are a small team paying for frontier-model generation and training runs out of pocket. A single high-reasoning dataset can cost more than a month of rent in tokens. If our models save you time, consider chipping in, or drop a distill request in the Discord and tell us what to build next.