SupraLabs Releases SupraElegans-500K: A C. elegans-Inspired Non-Transformer LLM

Dangerous_Try3619 · reddit · 2026-08-10

SupraLabs has released SupraElegans-500K, an experimental 500,000-parameter causal language model.

• Core Architecture: Abandons Transformers, attention mechanisms, and KV cache. It uses a sparse, signed, recurrent neural graph where context is carried by a persistent per-neuron membrane potential updated token by token.
• Inspiration: Loosely inspired by the C. elegans nervous system, featuring sparse connectivity, distinct neuron populations (sensory, interneuron, output), and a persistent recurrent state.
• Current State: This is a first working checkpoint to test if the architecture can do useful language modeling at a tiny scale. It is not tuned for quality or instruction-following and lacks a matched-parameter Transformer baseline.
• Features: Supports autoregressive generation and allows manual inspection or resetting of its internal recurrent state.

Original post →

More from Models

• DeepSeek's New Model Beats Nemotron3 Ultra With 4.2x Fewer Parameters — Scobleizer · 2026-08-10
• Sakana AI Validates Gemma 4-Based Conductor Model for Multi-Agent Orchestration — SakanaAILabs · 2026-08-10
• Snapcompact Halves API Costs by Rendering Text into Pixel-Font Images — DavidBennett__ · 2026-08-10
• Leak: OpenAI's Upcoming 10T Parameter 'Astra' Model Writes Like a Human — iruletheworldmo · 2026-08-10
• Discussion on Wan Animate 2 vs. SCAIL-2 Video Generation Models — Xxtrxx137 · 2026-08-10
• Rumors: Chinese open-weights models advanced by extracting reasoning traces from Claude Code and Codex — jxmnop · 2026-08-10

Models channel →