Diffusion-native tool-augmented reasoning

Continuous Interaction Diffusion

Run the 398.8M-parameter CID reference checkpoint and inspect how its visible Display changes across diffusion steps. CID maintains a latent Typed Cognitive Tensor (TCT), can express information needs during generation, call live retrieval, calculator, and symbolic math tools, and revise earlier output during iterative refinement.

This hosted Demo uses the portable reference runtime. Native deployments can use cid-engine for substantial execution acceleration.

CID Collection CID-v1-0.4B model Paper on Hugging Face Source code CID Engine

Try CID
Model
Compute

Tools

Deterministic numeric expressions

Exact algebra and calculus

Task-local document search and read
Examples
1 256
0.05 0.95
0.05 0.95
0.05 0.95
0.05 0.95

Result

Run at a glance
Readyrun status
—runtime steps
—wall time
—information needs

The visible Display is materialized repeatedly; changes below expose revision across model steps.
CID-v1-0.4B · 398.8M parameters · research prototype
Collection  ·  Model  ·  Code  ·  Paper