Model platform
Hollow Kiln
Status: In developmentA fully local platform for building, training, evaluating, and operating language models.
Kiln is Hollow’s model platform: repeatable local training, evaluation, inference, agent execution, experiment tracking, and traceable artefacts built around one set of canonical interfaces.
Hollow Kiln
Model platform
Build locally
A model platform, not a model-shaped demo
Kiln keeps model work local and inspectable, from tokenizer and dataset through training run, checkpoint, evaluation, inference, and agent execution.
- Repeatable training and evaluation workflows
- Traceable datasets, checkpoints, and model lineage
- A FastAPI backend, CLI, and React UI over the same interfaces
- Training and evaluation
- Local inference
- Experiment tracking
- CLI and web UI
Model families
Three directions, one careful platform
Kiln is developing Kindling, Ember, and Soot as distinct model families. The family pages explain the research direction without presenting checkpoints or pilots as finished releases.
- Kindling, Ember, and Soot have separate roles and lineages
- A trained checkpoint is not automatically a public release
- Weights, data policy, licence, and limitations stay explicit
- Kindling: small and efficient
- Ember: primary family
- Soot: open family
- Release status stays explicit
What’s inside
Everything around the model matters
- 01
Training
Local, repeatable runs with checkpoints and recovery.
- 02
Evaluation
Capability measurement and regression tracking against recorded baselines.
- 03
Inference
Model loading, generation, serving, and quantisation paths.
- 04
Agents
Tool-using model runs with explicit permissions and recorded context.
- 05
Lineage
Models carry the tokenizer, dataset, configuration, and run history that produced them.
- 06
Local first
The platform is designed around local experiments and inspectable artefacts.
Build models with a complete record of how they came to be
Hollow Kiln is in active development. Its model families are documented as research work, not public downloads.

