# llama.cpp update fixes Qwen2VL model scaling bug

> The b12385 release patches a scaling error that caused Qwen2VL to output garbage, improving its performance on CUDA and CPU.

Oossa · 2026-11-14 · https://oossa.com/en/llama-cpp-update-fixes-qwen2vl-model-scaling-bug

The open‑source llama.cpp library released version b12385 on November 14, 2026. The update addresses a bug in the Qwen2VL model where scaling factors for feed‑forward network layers were not applied. As a result the model produced nonsensical outputs, with perplexity around 1 e6 on CUDA and NaN on CPU. The fix passes the correct scales to the build routine, restoring normal operation.

## The facts

- Release version b12385 published on 2026-11-14
- Bug caused Qwen2VL perplexity ~1e6 on CUDA and NaN on CPU

## Why it matters

Developers using llama.cpp can now run Qwen2VL reliably, avoiding broken results in their applications.

## Sources & references

1. [ggml-org/llama.cpp b12385](https://github.com/ggml-org/llama.cpp/releases/tag/b12385) – llama.cpp, 2026-11-14

Last updated: 2026-11-14