Telos
The machine learning research library.
def complete(code):
draft→inspect→refine
deff(x):↵returnx+1
Autoregressive + bidirectional decoding
- Research
- Code language models
- Methods
- Autoregressive / masked diffusion
- Hybrid model
- COROSred
- Checkpoints
- 12.5M–100M class
- Tools
- telos-ml
Telos
Telos is a machine learning research library. Its current work includes small language models for Python code: autoregressive baselines, masked diffusion, and COROSred, which combines causal drafting and bidirectional refinement.
Its published preliminary reports investigate training token budgets, positional encoding, and differences between loss curves and code prediction capability. The project also maintains evaluations of other language models and the telos-ml Python package.
Publications
TelosPreliminary Report #2
télos (τέλος): Preliminary Report 2 — Capability Divergence, Positional Encoding, and the Limits of Masked Diffusion Scaling
August 2026 · Ivan Samuel
TelosPreliminary Report #1
télos (τέλος): Preliminary Evidence on Token-Budget Scaling in Masked Diffusion Language Models for Code Autocomplete
August 2026 · Ivan Samuel
