←
edge0(official)181 posts
edge0(official) avatar

edge0(official)

@Edge0_AI

edge0.ai | On-device foundation models. High intelligence, low memory, low power

◷ Joined September 2026▧ 22 Photos and videos
edge0(official) avatar
edge0(official)@Edge0_AI·Nov 12

A major turning point for our crew. Frontier-level agentic coding, now running fully on-device. Edge0-145B and Edge0-120B are available with open weights. Give them a try.

Introducing Edge0-145B and Edge0-120B, @Edge0_AI’s first device-native 100B-class foundation models with open weights. Cloud-level intelligence on your laptop. Free, offline, and fast enough for real vibe coding. - Smart: beats Claude Opus 4.6 on Terminal-Bench 2.1 (86.5% vs. 78.2%) - Light: needs just 29G memory, 1/3 of a standard setup (100G+) - Fast: decodes 75+ toks/s, prefills 1050+ toks/s on RTX Spark and Mac. - Steady: handles 128K–256K tokens of context without slowing down, even on long tasks Frontier-class agentic coding. No cloud API. No token bill. Your code and data never leave your machine. Open weights are live on @huggingface. Run it locally or plug it into your favorite coding agent.

0:45
edge0(official) avatar
edge0(official)@Edge0_AI·9h

Running a 145B-scale model traditionally involved multi-GPU server setups, often with H100s or H200s. Edge0-145B takes a different approach: a single 16–24GB consumer GPU, 32–64GB system RAM, and SSD offloading. The trade-off is throughput and concurrency. It's not a replacement for a multi-GPU server, but it makes local inference possible with much less hardware.

0:13
edge0(official) avatar
edge0(official)@Edge0_AI·Nov 17

11,000 PRs in a month is insane. Now imagine the token bill if every coding agent had to call a cloud API. This is why we're interested in running large models locally.

11,000 pull requests in one month. A revert rate around 0.5%. @poteto on the Grok Bot team's shipping pace at @spacexai. More contributors, daily stable releases, and an app that got faster. From AI Engineer Code Summit.

1:32
edge0(official) avatar
edge0(official)@Edge0_AI·Nov 12

35B parameters on a phone. Sounds like a lot to squeeze into RAM. Here's the thing: not all the experts need to be in memory at the same time. With Edge0-35B, we keep expert weights in local storage and bring in the ones needed for inference. We also use a trained prerouter to predict which experts are needed next, so we can start loading them while the model is still working on the current token. Still a 35B model. Just a smarter way to handle its memory.

edge0(official) avatar
edge0(official)@Edge0_AI·Nov 11

A robot doesn't need to write a 2,000-word essay while avoiding a chair. Smart glasses don't need to run a massive model just to understand “turn up the volume.” And your phone doesn't need a cloud-sized setup for every AI task. We talk a lot about how big a model can run on a device. Maybe we should talk more about what that device actually needs the model to do.

Post attachment
edge0(official) avatar
edge0(official)@Edge0_AI·Nov 10

Airplane mode has always meant giving up a few things. Local AI doesn't have to be one of them. Once Edge0-35B is downloaded and set up, you can disconnect from the internet and keep chatting with the model. A small change in your settings. A pretty big change in how AI works.

Post attachment