Models

August 3, 2026

Fabric 1.6 Latest

35B-parameter Mixture-of-Experts reasoning model with ≈3B active parameters per token, a hybrid linear-attention architecture, and a 262,144-token native context (extensible to 1,010,000). Built for agentic AI, coding, and enterprise workflows. Released under Apache 2.0.

Reasoning Model

July 27, 2026

Fabric 1.5 Instruct

0.7B parameter causal language model with Chunked Fabric Memory for 32K-token contexts. Fine-tuned for instruction following, coding, mathematics, and general chat. Released under Apache 2.0.

Instruct Model