Silicon Valley's New Secret: Chinese Base Models
From fine-tunes to founder stacks, the center of gravity is moving east.
Ovi's Tech Blog
Speeding up local LLM inference 2x with Speculative Decoding
How a small draft model can speed up LLM inference by 1.82× without sacrificing quality - benchmarking Qwen3-32B with speculative decoding
Open Weights, Borrowed GPUs
A practical guide to renting GPUs for running open-weight LLM models with control, privacy, and flexibility.
Harnessing GPT-OSS Built-in Tools
Learn how to properly set up vLLM with GPT-OSS built-in tools and integrate it with LibreChat to leverage powerful capabilities.