holleezy 🔴 ∎ 🦣

Self driving cars didn't start on every road. They started in one city, one set of conditions, with a safety driver. Same playbook here. Nolla's AI doctor starts with acne, in one state, with clinicians ready to step in. That's how you know it's real.

$25 free credits to builders for a model that replaces jev i promise you: the model routing your agent reads about 4 pages. then it makes the call. drex 1.5 reads 128k context. one pass. still pick an answer. #1 on the decision index. under 10B params. $0.04 per million input. faster than jev. it doesn't write a word. scores every option you give it. nothing to hallucinate. here is how you set it up: 1. Go to drex.nace.ai/invite/rx8cuzzc 2. Grab a free api key 3. point your agent at it Have fun tokenmaxxing.

StepFun has actually been cooking, and Step 5 preview might be their most interesting drop yet it scores 44 on the artificial analysis index. $0.71 per task. what stands out to me here is the intelligence-per-dollar, that's a solid spot to be sitting in 600b total / 27b active. sparse moe. 1m context. text + vision. i think the 1m context is a big deal for actually running agents $1 / $0.05 cached / $2.70 per 1m It is built for coding, long-horizon tasks and complex agent workflows. a lot of labs talk "agentic" but don't actually let you run the thing yourself, Its quite interesting that StepFun does going to be watching this one closely once the weights are out oct 15 platform.stepfun.ai