Koboldcpp v1.124 "aesthetic edition" released for local LLM inference
Fcking_Chuck · reddit · 2026-10-31
Koboldcpp, a popular runtime for running LLMs locally, has shipped v1.124, nicknamed the "aesthetic edition", built by 10 contributors. Details and downloads are on its GitHub releases page.
More from Infra
- Community asks: where is the Qwen 4 35B MoE as the Qwen 4 lineup nears — solyarisoftware · 2026-10-31
- Musk predicts orbital compute will outpace Earth's by end of 2028 as SpaceX targets 1M tons to orbit — XFreeze · 2026-10-31
- Lawyer runs DeepSeek-V4.1-Flash on 2x DGX Spark + 3090s: 2,618 tok/s prefill at 256K, 95.9% KLD accuracy, all data open — TumbleweedNew6515 · 2026-10-31
- CXMT validates G6 DRAM engineering samples in Q3, mass production expected within a year — zephyr_z9 · 2026-10-31
- AI scarcity moved to atoms: memory, GPU and wafer margins hit 58%-80% — sujingshen · 2026-10-31
- Anthropic's S-1 risk factors: compute tops the list at 13.4%, edging out regulation — ayushtweetshere · 2026-10-31