# llama.cpp 1.0 changes its API contract and tightens local inference server security

The November 13 llama.cpp and ggml 1.0 releases introduce semantic versioning for public interfaces, remove deprecated APIs and make several server behaviours safer by default. Native integrations and self-hosted agent endpoints need migration checks.

llama.cpp 1.0 removes 35 deprecated functions, adds FP8 KV caching and MoE expert caching, and restricts unauthenticated tools and private-address access. The new versioning promise helps downstream builders, but upgrading is not drop-in.

- Status: Active
- Published: 2026-11-15T10:09:34+13:00
- Updated: 2026-11-15T10:09:34+13:00
- Categories: Artificial Intelligence, Open Models, Inference & APIs
- Tags: local AI, open source, security
- Canonical HTML: https://beyondthe.news/dossiers/llama-cpp-1-0-ggml-semver-api-migration-server-security

## What changed

On November 13, 2026, ggml-org released llama.cpp v1.0.0 and ggml v1.0.0. The maintainers now promise major-version increments for public C API and ABI incompatibilities. This release removes 35 deprecated functions, llama_set_warmup and the defrag-thold option, renames a context-state function, and promotes NextN speculative decoding into llama.h. It also adds FP8 KV caching, host-memory MoE expert caching, new model architectures and an OpenAI-compatible text-to-speech endpoint. Server changes tighten handling of unauthenticated tools, local URLs and CORS proxy requests.

## Why it matters

Local AI products often embed libllama or expose llama-server to other services. A documented compatibility contract makes release planning more predictable, but the first 1.0 upgrade deliberately breaks older integrations and configuration flags. Server hardening also changes default behaviour for MCP/tool deployments and local media fetching. Builders should treat this as a tested migration, not merely an upgrade to a stable version number.

## Version numbers become a compatibility contract

The maintainer announcement defines major versions for breaking changes to the public C API, minor versions for compatible additions and backend/model support, and patch versions for fixes that preserve the API. The team also says API and ABI incompatibilities will receive major-version bumps. This is a forward-looking maintainer policy, not proof that all downstream packages are already compatible.

## The first 1.0 release removes old interfaces

The upstream release removes 35 deprecated llama.h functions plus llama_set_warmup and the defrag_thold context field and command-line flag. llama_set_embd_state_src becomes llama_set_ctx_state_src, and NextN speculative decoding moves from llama-ext.h to llama.h. In-process applications and scripts using these names must be rebuilt or changed; old binaries should not be assumed ABI compatible.

## Security defaults can interrupt agent and proxy configurations

The release hardens llama-server against server-side request forgery, DNS rebinding, CORS proxy access to local network addresses and unauthenticated tool exposure. Non-loopback tool/MCP setups now need authentication or an explicit opt-in to unauthenticated tools. Private-address media URLs require a separate opt-in. These changes reduce accidental exposure, but can disrupt previously working trusted-LAN workflows. Avoid restoring permissive defaults without understanding the trust boundary.

## New inference options are not universal speedups

The release adds OCP FP8 KV caches, a host-memory MoE expert cache with CPU-share controls, support for new model architectures and a /v1/audio/speech endpoint. FP8 cache footprint and model throughput depend on backend, context, hardware and workload. The release announcement contains capabilities, not an independently established across-the-board speed or cost gain.

## Key details

- Released November 13, 2026, as llama.cpp v1.0.0 and ggml v1.0.0.
- Major version bump promised for public C API/ABI incompatibilities; minor for compatible features, patch for fixes.
- 35 deprecated llama.h functions removed, along with llama_set_warmup and --defrag-thold.
- NextN speculative decoding API promoted to llama.h; state-source function renamed.
- FP8 KV caches and a configurable host-memory MoE expert cache added.
- llama-server hardens SSRF, CORS, DNS rebinding and tool authentication behaviour.
- New OpenAI-compatible /v1/audio/speech endpoint added.

## Builder takeaways

- Pin current builds and test v1.0.0 in staging before upgrading native integrations or service deployments.
- Search C/C++ code and launch scripts for removed function names and --defrag-thold, then rebuild against the v1.0.0 headers.
- Test llama-server tools and MCP behind non-loopback addresses with an API key; do not casually opt out of authentication.
- Audit any CORS proxy or private media URL workflow because local/LAN fetches may now be blocked.
- Benchmark FP8 and MoE cache options on actual model, quantization, backend and hardware before claiming memory or throughput savings.

## What to watch

- Downstream binding and package adoption of the 1.x interface contract.
- Further security guidance for MCP and tool-serving deployments.
- Independent performance measurements of FP8 KV caches and MoE expert caching across supported backends.
- Future minor/major release adherence to the published compatibility rules.

## Uncertainties

- The semantic-versioning contract is newly introduced; compliance across future releases is untested.
- Independent production performance and compatibility measurements are limited.
- The upstream notes describe changes but do not quantify the number of downstream integrations that will break.

## Sources

- [llama.cpp v1.0.0 release notes](https://github.com/ggml-org/llama.cpp/releases/tag/v1.0.0) — ggml-org / GitHub · primary · 2026-11-13T00:00:00+13:00. Authoritative release details for API removals, cache features, model support and server hardening.
- [llama.cpp v1.0.0: A stronger foundation for the local AI ecosystem](https://huggingface.co/blog/ggml-org/llama-cpp-v1) — ggml-org / Hugging Face · primary · 2026-11-13T00:00:00+13:00. Maintainer announcement and precise semantic-versioning policy.
- [llama.cpp 1.0 explained, FP8 KV cache and upgrading](https://www.systemdesign.academy/explained/llama-cpp-v1) — System Design Academy · independent · 2026-11-14T00:00:00+13:00. Independent practical cache test and upgrade discussion; hardware-specific results not generalised.

