Why hackers moved from llama.cpp to vLLM
1 hacker described this switch in public comments. Their words, with links to the originals.
What hackers said about the move
“I switched from llama.cpp to vLLM because of prompt cache bugs in qwen/gemma models”verdverm, 2026-06-03original comment ↗
Use llama.cpp? Get alerted when hackers criticize it.
Tell us the tools in your stack. When hackers start criticizing one of them, or moving to an alternative, you will hear about it first.
llama.cpp vs vLLM
| llama.cpp | vLLM | |
|---|---|---|
| Hackers who recommend | 1 | 1 |
| Hackers who criticize | 3 | 0 |
| Moved away from it | 2 | 0 |
| Moved to it | 1 | 1 |
Counts are distinct hackers in the sampled comments, not market share.
Criticism of llama.cpp
“llama.cpp master is still somewhat bad in Qwen 3.8 next performance”tarruda, 2026-09-09original comment ↗
“However, it seems there are still some issues that llama.cpp and stepfun need to work out before it's ready for everyday use.”sosodev, 2026-02-19original comment ↗
Praise for vLLM
No further praise for vLLM captured yet.