Ministral 3B (Mistral AI)

Ministral 3B sits in Mistral AI's edge-focused "Ministral" family, a line that the company first introduced through its October 2024 "Ministraux" announcement and has continued to iterate on for lightweight deployments. The documented generation is positioned as an "edge" model, a small footprint large language model that can run close to the user whether on-device or behind a low-latency gateway, while still offering a long 128k token context window that lets it handle extended documents, multi-turn conversations, and retrieval-augmented prompts. That combination of a compact profile with long-context behavior makes it useful for teams that want generative AI features without the cost or hardware demands of flagship frontier models.

Within Mistral's tiered platform the Ministral 3B generation has reached the end of its official lifecycle, with Mistral's documentation marking the variant as deprecated on the listed release date and pointing users to "Ministral 3 3B" as the supported successor for new integrations. For practitioners choosing between model options through an inference gateway, this means the most stable path forward is to migrate active projects to the replacement rather than initiating new workloads on the deprecated build. The lineage nevertheless remains a useful reference point for understanding how Mistral evolves its edge models, trading off parameter count and latency against context length and instruction-following breadth across successive 3B-class releases.