Ollama llama.cpp Errors: A Quick Lookup of Exact Messages and Fixes (2026)
A quick lookup of Ollama and llama.cpp error messages, grouped by pulling, loading and starting. Each row gives the exact text, the likely cause, whether we tested it, and a link to the full fix.
bufferAbout the author3 min read
This page collects the Ollama llama.cpp errors we have written full guides for, grouped by where they show up: pulling a model in Ollama, loading or running it, and starting llama.cpp. Match the exact message, read the likely cause, and open the full article for the fix. In the “Basis” column, “Tested” means the full article shows output we got on our own machine; other rows rely on the official docs or GitHub issues.
Ollama errors when pulling a model
| Error message | Likely cause | Basis | Full guide |
|---|---|---|---|
Error: pull model manifest: 412: The model you are attempting to pull requires a newer version of Ollama. | Your Ollama is older than the model needs. Upgrade, then check with ollama -v. | Tested (0.40.8 vs 0.40.12) | Fix the 412 |
... requires a newer version of Ollama that may be in pre-release | The model needs a pre-release build, so the latest stable installer is not enough yet. | GitHub issue | Install a pre-release |
Error: this model requires MLX support, but the MLX runtime is not available | For now this model is only supported on Apple hardware. The release notes say Linux and Windows support is coming. | Tested on Linux | Errors that look like 412 |
redirect target not allowed: … resolves to non-public 198.18.x.x | DNS sends the download host to a private or reserved IP, for example a proxy tool in fake-IP mode. It is a network problem, not a version problem. | Tested, plus a GitHub issue | Fix the DNS side |
Error: pull model manifest: file does not exist | The registry has no model with that name. In our test it was a model we had only created locally. | Tested | Our test output |
Ollama errors when loading or running a model
| Error message | Likely cause | Basis | Full guide |
|---|---|---|---|
error loading model: unknown model architecture: 'lfm3' | The GGUF uses an architecture that the llama.cpp build inside your Ollama does not know yet. Update Ollama, or run the file with a newer llama.cpp. | Tested (LFM3 GGUF on 0.40.12) | Fix unknown model architecture |
'nimble:latest' was downloaded in a format that is no longer supported. To re-download, run 'ollama pull nimble:latest' | Since 0.40.4, a decision model needs decision.type metadata. Re-pulling fixes official models; a third-party GGUF without that field cannot be fixed this way. | Tested on 0.40.12 | Fix the format error |
model "smoltest" is not a llama.cpp decision model | The model has neither decision.type metadata nor the decision capability, so Ollama refuses the decision request. | Tested | Why this happens |
llama.cpp errors when starting
| Error message | Likely cause | Basis | Full guide |
|---|---|---|---|
error: --model or -hf is required (older builds: error: --model is required) | llama-cli was started without a model. Add -m /path/to/model.gguf or -hf user/repo:quant. | Tested (build 11963) | Fix –model or -hf |
error while handling argument "-m": expected value for argument | -m was given but the path after it is empty, often because a shell variable is not set. | Tested | Our test output |
model name is missing from the request | llama-server was started with no model, so it runs in router mode and every API request must name a model. | llama.cpp server docs | llama-server router mode |
Ollama llama.cpp errors you can avoid before downloading
Some of these errors start with a model that is too new or too big for your setup. Before a large download, check the release notes on Ollama’s releases page and work out the memory you need.
- Our VRAM calculator estimates how much GPU memory a model needs from its size, quantization and context length.
- Our best local LLM guide suggests which models to try first for 8GB, 12GB and larger memory sizes.
- For OpenAI API, ChatGPT and Codex messages, see our OpenAI Codex error codes lookup.