Markdown source

Dense Slides: Minimax M2: Building the #1 Open Model – Olive Song, MiniMax

Source Video

Minimax M2: Building the #1 Open Model – Olive Song, MiniMax

Method

This deck is slide-only. The existing captured video frame set supplies candidate frames, then local OpenCV rejects sponsor/title/speaker-only frames, crops visible slide surfaces, deduplicates, and saves the cropped slide images.

Cropped Visible Slides

• Recreated text/layout view: open HTML recreation
• AI slide classifier: title_card confidence 0.89
• Text source: advanced OCR rapidocr-live/center-82/contrast.
• OCR decision: ready — Mixed title slide with embedded product/UI screenshots and small labels.

Slide text:

poneg

Models: CEw

Excel in all modalities M2 MiniMax 2.6 Speech 2.0 Music S

INTAMO Melean

URTLWK

Prosx!

Al-native Apps:

150m users globally Agent! Video Agentl

MedltrAesl witel yN MnMs Auls CeleLhefielpeeg HalsVidne Ape take

MINIMAX Intelligene

E/LEAD MINIMAXM2

SENTED 8Y eenMinc OLIVE SONG /Senior Research Engineer MINIMAX

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.99
• Text source: agent_vision.

Slide text:

MiniMax-M2: Open-Weight. Coding-First. Best-in-Class.

~10B activated parameters, open-weight

Agentic-by-design for coding and workplace tasks

Fast, cost-efficient, and ready to scale

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.97
• Text source: agent_vision.

Slide text:

Training Shapes M2 Behavior for Developers

Model Characteristics

Training

1. Coding & Dev Experience

Scaled Environments & Experts

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.98
• Text source: agent_vision.

Slide text:

Interleaved Thinking Enables...

1. Adaptation to Environment Noise

2. Focus on Long-Horizon Tasks

• Recreated text/layout view: open HTML recreation
• AI slide classifier: content_slide confidence 0.98
• Text source: advanced OCR rapidocr-live/bright-screen/opencv-adaptive.
• OCR decision: ready — Multi-column slide with multiple embedded UI screenshots and a QR code; OCR is better for the small screenshot text.

Slide text:

M2 Scales and Collaborate in Multi-Agent Systems

Research Agent: Web Development Agent Report Agent.

Small & Cost-effective to run long agentic tasks

MI MINIHAX MiniMax Agent Intelligcnco wlth Everyero

Classification audit: raw/sources/slide-ai-classification/dense/lY1iFbDPRlw/audit.json