
turbo-tan/llama.cpp-tq3
Read this first. The README says the TQ weight formats are now developed in tqLLM, where new features land first; this fork stays for llama.cpp users.
Adds TQ3 weight formats to llama.cpp and an FP4 prompt path on Blackwell GPUs. It is a fork of ggml-org/llama.cpp.
Benchmarks
- On GB10 / DGX Spark (sm_121) with Qwen3.6 27B MTP (TQ3_4S), it reads prompts at 919.63 tok/s, against 359.94 tok/s for FP4 cache path off (2.6×). Source Note: Prompt processing at 2048 tokens with -ngl 99. The FP4 path is a runtime option on Blackwell GPUs; the model file stays TQ3_4S.
Listing
Listed: Passed all seven checks.
Security review
Not reviewed yet.
Install
To install it, follow the project's own instructions on GitHub. The engine runs with your user's permissions. macOS doesn't sandbox it, so install only engines you trust.
Something wrong or out of date? Write to help@kalsa.io, or ask us to remove this page.