This looks great but I'd want to make sure to do real work as the benchmarks for multiple reasons, and so want this to take part in a CI/CD instead of doing synthetic benchmark. Have you tried…
An agent skill for hill-climbing local LLM speed, not a list of flags. It computes the ceiling first, changes one thing at a time, and logs the failures too. On an Arc B390, 1-user decode went 43.6…
Before tuning a local model, it helps to know how fast the hardware can go at all. We ran the calculator in a new open agent skill and got that ceiling in a second. It also showed that chat speed…
691 stars in a day: an agent skill that hill-climbs your local llm server's speed. compute the physics ceiling first, change one variable at a time, gate every result on a long-context stability…
An agent skill that makes local LLMs run faster just went open source: 380 GitHub stars in 13 hours. You point Claude Code at your model server. It works out the top speed your hardware allows,…
Find the local LLM that actually runs and performs best on your hardware. Ranked by real, recency-aware benchmarks, not parameter count. One command, run it instantly.
🏋️ A unified multi-backend utility for benchmarking Transformers, Timm, PEFT, Diffusers and Sentence-Transformers with full support of Optimum's hardware optimizations & quantization schemes.