REDDIT · REDDIT// 4d agoOPENSOURCE RELEASE

Horus 1.0 debuts as Egypt's open model

TokenAI is launching Horus 1.0, an open-source 4B text model series trained from scratch in Egypt, with the first release shipping in multiple compressed variants for different hardware tiers. The model targets Arabic and English use cases, supports 8K context, and is wired into the neuralnode Python framework.

// ANALYSIS

This is a meaningful regional model launch: the important part is not just the “first in Egypt” framing, but that TokenAI is packaging a scratch-trained open model with practical deployment options developers can actually use. The benchmark claims look strong on paper, but they should be treated as vendor-reported until independent evals confirm them.

–Open weights plus multiple quantized builds make the release more useful than a single flagship checkpoint
–The Arabic-English focus matters if TokenAI can sustain quality beyond headline benchmark gains
–Neuralnode integration lowers the friction for trying the model in Python workflows
–If the MMLU and MMLU Pro results hold up, Horus 1.0 becomes a serious small-model contender rather than a symbolic launch

// TAGS

horustokenaillmopen-sourcereasoningbenchmark

DISCOVERED

4d ago

2026-04-08

PUBLISHED

4d ago

2026-04-08

RELEVANCE

9/ 10

AUTHOR

assemsabryy