@intellectronica As an OpenClaw orchestrator, on max, I feel it is about on par with GPT-5.5: Medium while orchestrating a swarm. It does seem to be very fast even on Max. I may config the swarm agents to GLM-5.2 for speed/$$$, but I will leave my orchestration with GPT-5.5: High (still better).
@thsottiaux GPT-5.5 HIGH via OpenClaw has been faster and more capable since this morning than it has ever been. Anyone else experiencing speed/quality improvements since about 8am ET? Feels like an upgrade 🔥
@mark_k@Kimi_Moonshot Wanted to stay on 🇺🇸 models. But... I've been using coding loops and Final-QC feedback loops with my orchestrator and coding swarm agents using the latest from Kimi, and, my results have been excellent. Kimi is closing the gap... nice release @Kimi_Moonshot
@mark_k Need to optimize power/hardware usage... models (re)building themselves is likely the only chance that ASI will be for everyone (exponential capability without cost exponential++).
@mark_k@sunnysampla_@OpenAI Yep - and it is fast and can be done with oauth -> x_search in openclaw/hermes (use grok-4.1-fast so you don't burn your supergrok quota .... don't need 4.3 to scrape X).