Edge0 launches open-weight 145B/120B device-native models, 145B beats Opus 4.6 on Terminal-Bench
hey_abusiddik · x · 2026-11-13
Edge0 AI introduced Edge0-145B and Edge0-120B, its first device-native 100B-class foundation models with open weights, targeting cloud-level agentic coding fully offline on a laptop. Claimed specs: 86.5% vs. Claude Opus 4.6's 78.2% on Terminal-Bench 2.1, just 29G memory (about 1/3 of a typical setup), 75+ toks/s decode and 1050+ toks/s prefill on RTX Spark and Mac, and steady 128K–256K context handling. Weights are live on Hugging Face — vendor-claimed numbers, worth independent verification.
Related event: Edge0 Releases Open-Weight 145B/120B On-Device Models(4 posts)→
More from Infra
- Microsoft Cloud tops $63B as AI buildout keeps scaling, says Nadella — thedealdirector · 2026-11-14
- Meta exec: AI feels weightless, but its infrastructure has a real physical footprint — moniquejmorrow · 2026-11-14
- NVIDIA touts SGLang partnership: 5 GPU generations, 60+ day-zero model launches — NVIDIAAI · 2026-11-14
- Qwen3.5-9B runs offline on a stock PS4 at 39 tok/s using its own MTP head — leonardo_alegria · 2026-11-14
- Claude's Microsoft 365 connector down ~20 hours, exposing agent dependency risks — krishnan · 2026-11-14
- llama.cpp 1.0 commits credit a locally-run DeepSeek on Mac Studio as coding assistant — krishnan · 2026-11-13