Bittensor subnet 92 releases REAX 1.3, an open-source decision model blending miner models
const_reborn · x · 2026-11-06
REAX is a decentralized lab on Bittensor subnet 92: miners train decision models in open competition, and releases build on the best entries. REAX 1.3 is now open-sourced on Hugging Face under Apache-2.0, claiming the top spot on Jev-style decision-model leaderboards (74.0, tied #1 in the shared benchmark chart).
- Two parts run in parallel on a single RTX PRO 6000 (96GB); per-option probabilities are averaged 50/50 and returned via a Jev-compatible interface.
- Part G blends miner model uid 18 (Gemma-lineage 31B fine-tune) with PelaAI/KnowLine-31B-Gen1; Part Q blends miner uid 83 (Qwen3.5-27B lineage) with vllm-sr/d3-ultra (27B).
- Inference runs via the MIT-licensed REAX Engine (reax13 mode); needs Linux, CUDA 13/SM120, Docker, 400GB disk and 64GiB RAM (128GiB recommended for FP8 weight loading).
Related event: REAX 1.3, trained on Bittensor subnet in four weeks, tops HF Decision Index(5 posts)→
More from Infra
- Ron Baron backs Musk's space data centers, targeting 2027-2028 solar-powered orbital launches — elonmusk · 2026-11-06
- Turso launches Singapore datacenter, devs cheer the new Asia-Pacific region — glcst · 2026-11-06
- Ex-Meta employee estimates Muse costs $150-$300 per user monthly — RihardJarc · 2026-11-06
- llama.cpp PR adds ClassLRU MoE cache: +6-7% decode speed and 8% fewer uploads vs SLRU — pmttyji · 2026-11-06
- JPM's Global Cash Trading Co-Head: AI's Circular Impact on Markets Was Under-Scrutinized — ZeroStateReflex · 2026-11-06
- LLVM Dev Meeting 2026: Building durable open-source infrastructure in the agentic era — clattner_llvm · 2026-11-06