How to run MiniCPM-o 2.6 locally

OpenBMB · Omni understanding + speech generation · 8B params · MiniCPM Model License (code Apache-2.0) (commercial OK)

GPT-4o-style 8B omni model that takes images, video and audio and supports real-time bilingual speech-to-speech conversation with voice cloning and configurable voices. Free commercial use is permitted after completing OpenBMB's registration form.

What it costs to run — $0 markup

On your own machine — $0. Runs free if you have about 9.0 GB of VRAM (~17-19GB in BF16; the official int4 build runs real-time speech and vision in roughly 7-9GB, suitable for a single mid-range GPU.), via Transformers (also llama.cpp / vLLM).
Rent a GPU — from $0.26/hr. Fits RTX 3090 24GB at the direct vendor price ($0 markup) — pay only for the minutes you generate.
Download the weights — free. Open weights at openbmb/MiniCPM-o-2_6.

Note: generative-media models are billed per image / per second / per minute on hosted services — not per token. Running locally or on your own rented GPU is usually far cheaper and keeps your data on your machine.

Key facts

Does	Visual understanding, Video understanding, Audio understanding, Speech recognition, Text → speech, Speech → speech, Ocr
VRAM to run	~9.0 GB (~17-19GB in BF16; the official int4 build runs real-time speech and vision in roughly 7-9GB, suitable for a single mid-range GPU.)
Download	~17 GB
Parameters	8B
License	MiniCPM Model License (code Apache-2.0) (commercial use OK)
Run with	Transformers (also llama.cpp / vLLM)

Get MiniCPM-o 2.6 on Hugging Face →

More multimodal / omni models

Qwen2.5-Omni-7B — Omni understanding + speech generation, ~24 GB VRAM
Janus-Pro-7B — Unified understanding + image gen, ~16 GB VRAM
OmniGen2 — Any to image (gen + edit + in context), ~17 GB VRAM
BAGEL-7B-MoT — Unified understanding + image gen + edit, ~24 GB VRAM
Emu3.5 — Any to any world model (gen + edit), ~48 GB VRAM
Emu3-Gen — Next token any to any generation, ~18 GB VRAM

Browse: all media models · chat / LLM models