El informe completo. Puntuación, pilares, desglose matemático, riesgo, escenario, señal y notas.
Investigación, no asesoramiento financiero. Lee los términos.
Acerca de τemplar
Análisis del 12 nov 2026En inglés Traducción pendiente: este análisis se muestra en inglés.
τemplar (SN3) is a decentralized AI pre-training protocol built as Subnet 3 on the Bittensor network. The platform coordinates globally distributed GPUs over the open internet to train large language models without centralized data centers.
Resumen completoMostrar menos
In March 2026, the subnet completed Covenant-72B across more than 70 permissionless nodes. That run trained a 72.7-billion-parameter LLaMA-style model on 1.1 trillion tokens using SparseLoCo gradient compression.
On April 10, 2026, founding developer Covenant AI announced its withdrawal from Bittensor over network governance and emission disputes. The original Covenant-72B codebase and Apache 2.0 model weights remain open on GitHub and Hugging Face.
Subnet 3 tokenomics cap total supply at 21 million SN3 tokens under the Bittensor dynamic TAO framework.
