Install Ollama 0.13.3 or newer and run ollama run devstral-small-2, the 24B model Mistral AI built for coding agents. Ollama lists it as a 15GB download that reads text and images and supports tools, and its model card gives the licence as Apache 2.0 and a 256k context window. The earlier Devstral, ollama run devstral, is a 14GB text-only 24B download, and Devstral 2, ollama run devstral-2, is a 123B model and a 75GB download. Mistral's card calls the Small 2 light enough for a single RTX 4090 or a Mac with 32GB of RAM. None of the three is in Grux's catalogue, so this site rates none of them on any Mac and quotes download sizes only, which are usually smaller than the memory a model needs while running.
| Model | Command | Ollama lists | Input |
|---|---|---|---|
| Devstral Small 2, 24B | ollama run devstral-small-2 | 15GB download; needs Ollama 0.13.3 or newer | Text and image |
| Devstral, 24B (the earlier one) | ollama run devstral | 14GB download | Text |
| Devstral 2, 123B | ollama run devstral-2 | 75GB download | Text |
Devstral Small 2 also has a cloud tag on Ollama, which runs on Ollama's servers; the commands above run locally.
Grux's catalogue lists no Devstral model, so the Mac picker does not rate one, and this page writes no rating for it on any memory size. The only sizes here are Ollama's download sizes. A 75GB download is larger than the memory of any Mac up to 64 GB, which is arithmetic and not a rating. For coding models that are in the catalogue, see best local LLM for coding on a Mac.
Ollama's page for each Devstral lists launch commands for Claude Code, OpenCode, Hermes Agent and OpenClaw. This site's pages on Claude Code with a local model and on coding agents with a local model give those steps with Ollama's own advice, including the context window it asks for.
| Claim | Source |
|---|---|
| Devstral Small 2: 24B, 15GB, text and image, tools, Apache 2.0, Ollama 0.13.3 or newer | Ollama library, devstral-small-2 |
| Devstral: 24B, 14GB, text | Ollama library, devstral |
| Devstral 2: 123B, 75GB | Ollama library, devstral-2 |
| Apache 2.0, 256k context, the RTX 4090 and 32GB Mac statement | Hugging Face, Devstral Small 2 24B Instruct model card |
ollama run devstral-small-2, a 15GB download of the 24B Devstral Small 2.