GitHub radar
MiniMax M3: 1M context coding model via Ollama
MiniMax's flagship model for coding and agentic tasks is now accessible via Ollama. It holds up to 1M tokens of context, reads images natively, and scores 83.5 on BrowseComp — above Claude Opus 4.7.
MiniMax M3 is MiniMax's flagship model for coding and agentic tasks, now accessible through Ollama's cloud platform. The context window reaches up to 1 million tokens (guaranteed minimum 512K) — enough to load an entire medium-sized codebase, long conversation history, or extensive documentation into a single request without truncation. The model is natively multimodal: it processes text and images in one unified pipeline, so you can attach a screenshot alongside text and it understands both together. On BrowseComp — a benchmark measuring how well a model autonomously finds and synthesizes information from the web — MiniMax M3 scores 83.5, compared to 79.3 for Claude Opus 4.7. This is a cloud-only model: it runs on MiniMax's servers and is available through Ollama at $0.60 per million input tokens and $2.40 per million output tokens. The weights are publicly released, but running them locally requires data-center-scale hardware.
Why a vibe-coder should care
If your project has grown large enough that context limits force you to split every request, M3's million-token window changes the workflow: send the whole thing at once. The multimodal input lets you drop in screenshots of errors or UI mockups directly — no need to describe them in words.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Connect MiniMax M3 via Ollama: ollama run minimax-m3 — requests go to MiniMax's cloud. After connecting, show me how to send a request with an image attached, and how to confirm that the million-token context window is actually in effect.
Cloud model — runs on MiniMax servers, can't be installed locally. Available via Ollama: $0.60 per million input tokens, $2.40 per million output tokens.
Open on Hugging Face▌ More finds