ggml llama.cpp releases
new versionggml llama.cpp is tracked under self-hosted apps & servers. Its latest version, 0.9.0, shipped on Nov 4, 2026. Firmwarely has recorded 6 versions since Sep 14, 2026.
- Latest version
- 0.9.0
- Released
- Category
- Self-hosted apps & servers
- Last checked
How to update ggml llama.cpp
- Read the release notes first — self-hosted projects break config and schemas between versions more often than appliances do.
- Docker: pull the new image tag and recreate the container (`docker compose pull && docker compose up -d`).
- Package or bare metal: update through your distribution's package manager, or replace the binary from the releases page.
- Back up the config and database directory first. For most projects a downgrade is not supported once the schema has migrated.
Always install releases from the project itself. Firmwarely links out; it never hosts downloads.
What changed
Overview: This release adds OCP FP8 E4M3 quantization support, a new flash_attn_ext_rows op that makes the unified KV cache efficient for multi-sequence inference, and a generic batched tensor copy API. Four new architectures are supported (Clover, Limite, Metis-1.6 and Mistral Large 4), audio gene…
Version history
| 0.9.0 | |
| 0.8.0 | |
| 0.7.0 | |
| 0.6.0 | |
| 0.5.0 | |
| 0.4.1 |
FAQ
What is the latest version of ggml llama.cpp?
Version 0.9.0, released Nov 4, 2026. This page is refreshed every hour from the project's release page.
How do I update ggml llama.cpp?
See the step-by-step above. Pull the new image or package and restart the service. Read the release notes and back up the data directory first; most projects can't downgrade once the database has migrated.
How does Firmwarely know when there's a new version?
Every hour we read the project's official release page for ggml llama.cpp and record the version, date and changelog. Subscribers watching it get an email when it changes: the next morning on Pro, in the Monday roundup on the free plan.
Is this an official ggml page?
No. Firmwarely is independent. Project and product names belong to their owners; always install releases from the project's own source.
Get alerts for ggml llama.cpp
One email when a new release or a security fix ships: weekly and free for up to 3 devices, daily on Pro.