Sun, May 31, 2026
A quiet Sunday on the wire. Two of three releases are AI infrastructure plumbing. llama.cpp patches tensor parallelism granularity for Qwen 3.5 and 3.6 models running on three GPUs. koboldcpp adds experimental continuous batching to handle parallel text generation requests instead of queueing. One Nostr maintenance release rounds out the day.
Nostr
- Jumble 26.5.7. chore: release v26.5.7 Co-Authored-By: Claude Opus 4.7 (1M context)
AI
- llama.cpp b9434. TP: fix granularity for Qwen 3.5/3.6 + 3 GPUs ( 23843) TP: fix granularity for Qwen 3.5/3.6 + 3 GPUs fix afmoe TP macOS/iOS: - macOS Apple Silicon (arm64) - macOS Apple Silicon (arm64, KleidiAI enabled) DISABLED - mac…
Matters only if running Qwen 3.5 or 3.6 models with tensor parallelism across exactly three GPUs.
- koboldcpp 1.114. koboldcpp-1.114 https://github.com/user-attachments/assets/39732143-2f0b-4546-bc37-f66ba071c4fd - NEW: Experimental parallel text generation requests (continuous batching) is now optionally supported - Normal text gen…
Parallel request handling is opt-in; no action required unless workload benefits from concurrent batching.
Read this brief on the web: https://freedomtech.news/posts/2026-05-31-bitcoin-daily-brief/
No comments yet.
Write a comment