# Ggml-org Merges hc Ops into qwen4exp Graph in llama.cpp

Digest AI · Generative AI & Models · published 2026-09-16T09:51:41Z

Canonical: https://digestai.news/story/ggml-org-merges-hc-ops-into-qwen4exp-graph-in-llama-cpp

## Summary

A recent pull request (#28901) in the open‑source llama.cpp repository has added new high‑capacity (hc) operations to the qwen4exp graph, targeting both CPU and CUDA backends. The changes fuse a per‑element gate into the hcpre stage and combine an identity operation into hcpost, streamlining the graph’s execution path.

The merge was approved by key contributors feal87, pwilkin, and ggerganov, and was assisted by the deepseek‑v4‑flash team. While the commit itself is a modest code tweak, it reflects the active community effort to optimize LLM inference pipelines.

These updates may reduce computational overhead and improve throughput for models that rely on the qwen4exp graph, making the open‑source stack more efficient for developers and researchers alike.

## Key points

- Pull request #28901 merges hc pre/post ops into qwen4exp graph for CPU and CUDA.
- Key contributors: feal87, pwilkin, ggerganov.
- Change improves gate fusion and identity combination in llama.cpp.

## Why it matters

The update streamlines internal operations for the qwen4exp graph, potentially boosting inference speed and reducing memory overhead on both CPU and GPU. It demonstrates ongoing community‑driven optimizations in open‑source LLM tooling.

## Sources

1. [qwen4exp: add hc ops by am17an · Pull Request #28901 · ggml-org/llama.cpp](https://github.com/ggml-org/llama.cpp/pull/28901) (github.com, 2026-09-16, primary source)

Part of the developing story: [Open Source AI Infrastructure and Medical Model Expansion](https://digestai.news/thread/ggml-org-merges-hc-ops-into-qwen4exp-graph-in-llama-cpp) (4 stories)

## Cite

Digest AI, "Ggml-org Merges hc Ops into qwen4exp Graph in llama.cpp", 16 September 2026, https://digestai.news/story/ggml-org-merges-hc-ops-into-qwen4exp-graph-in-llama-cpp

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/ggml-org-merges-hc-ops-into-qwen4exp-graph-in-llama-cpp.json