Activity and trending
Stars, forks and merged pull requests each day or month, lined up with the repository's rank on Trendshift Trending
Qwen3.8-Flash-Next at 100+ tok/s on one NVIDIA DGX Spark (GB10), ~2× the typical ~50 tok/s recipes. 101 tok/s prose, 133 code, 250 tok/s with 4 requests, plus image input. Hibana engine, MTP speculative decoding, Docker Compose, OpenAI/Anthropic API.
Stars, forks and merged pull requests each day or month, lined up with the repository's rank on Trendshift Trending
Single DGX Spark owners: Qwen3.8-Flash-Next at 101 tok/s on prose (82 at temp 1.0). 2.2× the typical recipe in the same test. Most Flash-Next speed posts lead with code. The 101 here is prose, the…
@iamprestonlogan · x.com