json is so token inefficient it hurts these days man, these braces and quotes are costing me real $$
LLMS
-
Testing concurrency, nvfp4, and TP performance across BF16, FP8, nvfp4
By
–
still need to test it in concurrency + try it in nvfp4 hopefully will be able to compare perplexity in bf16 / fp8 / nvfp4 as well as tp performance jumps from 2 -> 4 -> 8 nodes across all 3
-
MiniMax M2.7 on 4x DGX Sparks with vLLM and OpenCode
By
–
MiniMax M2.7 at home running on 4x DGX Sparks
— Ahmad (@TheAhmadOsman) 12 avril 2026
vLLM serving full BF16 weights, 200k context
OpenCode having the model monitor its own hardware and report thermals, tokens/sec, TTFT, and other runtime stats in real time
What benchmarks / workflows / things do you wanna see next? https://t.co/iCR8UsLs8E pic.twitter.com/I04c7j6UkgMiniMax M2.7 at home running on 4x DGX Sparks vLLM serving full BF16 weights, 200k context OpenCode having the model monitor its own hardware and report thermals, tokens/sec, TTFT, and other runtime stats in real time What benchmarks / workflows / things do you wanna see next?
-
Summarized thinking traces useful despite trade secret internal reasoning
By
–
(thinking traces are always summaries of the internal reasoning, since the actual internal chain of thought is a trade secret, but that doesn't mean they aren't useful when summarized, especially when those summaries contain the tools used by the model and web searches it made)
-

Lightweight AI Agent Processing Memory Continuously 24/7
By
–
An AI Agent that runs 24/7 as a lightweight background process, continuously processing, consolidating, and connecting information. Just an LLM that reads, thinks, and writes structured memory. Built with Gemini 3.1 Flash Lite.
-

OpenClaw GPT-5.4 Update: Finding the Sweet Spot for Daily Use
By
–
OpenClaw with GPT-5.4 feels noticeably better with the new update. GPT-5.4 (thinking=high + fastmode=true) seems like the sweet spot. Better judgment, less back-and-forth, still fast enough for daily use.
-

AI models debate power usage of local OpenClaw on Mac
By
–


And if you are interested in the answer, Claude and ChatGPT 5.4 Pro think running OpenClaw with local inference on your Mac uses more total power. Gemini doesn't think so (but appears to have not thought about that too hard)
-

ChatGPT best at thinking traces, Claude good, Gemini weak
By
–


Currently, ChatGPT has the best way of viewing thinking traces, a short summary of steps in the main window, and a detailed audit in the sidebar if you want it Claude does almost as well, but more summarized and harder to see calculations and code Its a big weak spot for Gemini
-
Comparing 26B A4B and 31B Model Performance
By
–
Have you tried the 26B A4B one? I've been impressed by it, I've not compared it to the 31B yet though