← 回總覽

Bittensor (TAO) 生态系统子网分析

📅 2026-03-16 22:52 0xSammy 人工智能 2 分鐘 2296 字 評分: 84
Bittensor TAO 子网 分布式训练 GPU 市场
📌 一句话摘要 一份关于 Bittensor 生态系统的深度分析补充,重点关注 Templar、Grail 和 Basilica 等各子网的表现与角色。 📝 详细摘要 这条推文是主综述的技术补充,引用了对 Bittensor ($TAO) 生态系统的详细分析。它拆解了“旗舰级”子网,包括用于去中心化训练的 SN3 (Templar)、用于训练后处理的 SN81 (Grail) 以及用于 GPU 计算的 SN39 (Basilica)。报告强调了大规模无许可分布式训练的成功证明,以及医疗和存储等专业子网的增长。 📊 文章信息 AI 评分:84 来源:0xSammy(@0xSammy) 作者

Title: Bittensor (TAO) Ecosystem Subnet Analysis | BestBlogs.dev

URL Source: https://www.bestblogs.dev/status/2033557026060534193

Published Time: 2026-03-16 14:52:31

Markdown Content: Big weekend for $TAO, up 41% in 7 days The catalyst for the ecosystem was SN3 @tplr_ai announcing results from Covenant-72B, the largest decentralised LLM pre-training run in history. 72B params, 1.1T tokens, no centralisation.

Templar is up 178% adding $50mm to their market cap.

Don't sleep on SN81 Grail and SN39 Basilica, led by the same team as Templar.

Subnet ecosystem just crossed TOTAL MC $1.1B:

  • SN3 @tplr_ai at $79.6mm MC (+163%): Decentralised LLM pre-training, the flagship of Bittensor's distributed training thesis. Most liquid subnet at $30mm last 24h. First proof permissionless distributed training CAN WORK at frontier scale.
  • SN81 @grail_ai, at $20.1mm MC (+81%): Post training subnet from Templar team, different architecture; different coordination mechanism for the same goal of decentralised model training
  • SN39 @basilic_ai at $21.9mm MC (+42%): GPU compute marketplace with transparent pricing; Same team as Templar / Grail. Basilica procures the compute, Templar pre trains and Grail does the post training. $7.8mm last 24h.
  • SN68 @metanova_labs at $20.6mm MC (+97%): Decentralised drug discovery, molecular simulation and protein folding on Bittensor. One of the few subnets applying Bittensor novel mechanism to healthcare use cases. $6.2mm last 24h
  • SN93 @Bitcast_network at $14.5mm MC (+86%): incentivises creators to produce educational Bittensor content. Recently launched no-code miner setup expanding the contributor base. Covenant-72B + NemoClaw drove a surge in Bittensor awareness and new creators flooded in
  • SN75 @hippius_subnet at $38.1mm MC (+42%): decentralised cloud storage for Bittensor; every subnet that stores model weights or datasets needs persistent storage, when the ecosystem grows, the storage layer re-rates
Follow @0xSammy and our team at @KhalaResearch for more research on TAO.

查看原文 → 發佈: 2026-03-16 22:52:31 收錄: 2026-03-17 02:01:04

🤖 問 AI

針對這篇文章提問,AI 會根據文章內容回答。按 Ctrl+Enter 送出。