b10067
ggml-org/llama.cppb10067Jul 18, 2026by github-actions[bot]1 days after b10066
What changed since b10066
Fix quantization crash for DeepSeek-V4 routing tables.
Bug fixes
- Fix quantization crash for DeepSeek-V4 routing tables (issue #25754).
- Fix OpenCL MoE kernel loading (issue #25797).
Upgrade notes
The llama-quantize tool now correctly skips the ffn_gate_tid2eid index table during quantization, preventing crashes on DeepSeek-V4 models. No configuration changes are required.
Analyzed Oct 16, 2026
Upgrading?
Release notes summary
Fix for quantization of DeepSeek-V4's routing table tensor.
Key Highlights
- Exclude i32 ffn_gate_tid2eid routing table from quantization
- DeepSeek-V4's routing table is an index table, not weights
- Fixes issue #25754
Release Notes (excerpt)
<details open> llama-quant : exclude i32 ffn_gate_tid2eid routing table from quantization (#25787) DeepSeek-V4's ffn_gate_tid2eid tensor is an i32 token-id -> expert-id index table, not weights. It was never added to the name-based exclusion list alongside ffn_gate_inp.weight, so llama-quantize tries to quantize it and fails since i32 cannot convert to a float type. Fixes ggml-org/llama.cpp#25754 Signed-off-by: Yash Raj Pandey <yashpn62@gmail.com> </details> **Website:**…
Published by ggml-org on GitHub.
Read the full release notesRelated Repositories
:electron: Build cross-platform desktop apps with JavaScript, HTML, and CSS
Godot Engine – Multi-platform 2D and 3D game engine
ClickHouse® is a real-time analytics database management system
Self-hosted game stream host for Moonlight.
DuckDB is a high-performance analytical database system