whereismyai
Speck
Speck: Cognitive runtime makes small language models more capable.
Speck is a cognitive runtime that makes small language models more capable by moving cognition out of the prompt and into persistent, deterministic software. It separates cognitive state from model inference, allowing the model to participate in cognition without owning it. Speck contains runtime mechanisms for processes such as persistent memory, attention, evidence, and planning. It is designed to work with locally hosted models and services, and is built on the Genesis runtime. Speck's architecture is functional, but it is still an active experimental project. The project includes benchmarks, probes, and ablation infrastructure to test the effectiveness of its cognitive mechanisms.
Speck at a glance
- Best for: Software engineer, Researcher
- Runs on: Node.js
- Pricing: Free
- Released: October 2026
- Category: AI research tool
- Website: github.com/doctarock/speck
Key features
- Persistent Memory
- Attention Mechanism
- Evidence Tracking
- Planning
- Metacognition
Use it for
- Improving small language models
- Enhancing cognitive capabilities
- Supporting local inference
Alternatives to Speck
- Overview — Learn language models by running hands-on experiments without writing scripts
- Every number in a tiny GPT — Inspects every lookup, matrix multiplication, and normalisation in a live transformer.
- VLMEvalKit — Evaluate large vision-language models across numerous benchmarks with a single command.
- LiveBench — Compare model performance and operational costs without contamination.
- Lighteval — Evaluates LLMs across multiple backends with thousands of built-in tasks.
- Vegaml — Runs locally on your hardware with calibrated probabilities and conformal answer sets.
- Lune — Accesses trusted knowledge from top conference papers for AI agents.
- Jevon Tane — It provides a rigorous, pre-registered framework for AI agents to conduct science together.
All Speck alternatives