2026-05-22

MiniMax at the WaytoAGI Global AI Conference

On April 8th, we joined WaytoAGI at Happo-en in Tokyo for a global AI conference that brought together founders, builders, and operators from across the Chinese and global AI ecosystems. It was a chance to share how we think about multimodal intelligence, hear what teams in Japan and across Asia are building.

On stage: Multimodal Agents and a New Model of Human–AI Collaboration

Cherie from MiniMax gave an afternoon talk on how our product matrix sits on top of a single, self-developed model stack — and what that unlocks for the people building with us.

Cherie on stage at WaytoAGI Global AI Conference

A few of the ideas she shared:

One model stack, many products. MiniMax's product matrix — Talkie (one of the largest generative AI mobile applications globally, per a16z), MiniMax Agent, Hailuo AI, and our audio and music platforms — is powered by the same in-house model stack, with the same capabilities exposed to external partners through an open API.

Where the models are today. Speech 2.8 supports 40 languages and 90+ dialects, with Japanese output reaching near-native quality. Hailuo 2.3 has built a substantial user base in Japan and meaningful penetration within the Japanese AI animation creator community. M2.7 is one of the default base model options on OpenClaw — described by OpenClaw's founder as the best open-source model he has used, at 5% of the cost of OpenAI and Anthropic and three times the speed.

Where we are as a company. Listed in Hong Kong with a market cap of over $50B, 100,000+ enterprise customers, 70%+ of revenue from overseas markets, and users across 200 countries and regions.

Roundtable: Beyond Text — How AI Is Reshaping Creative Workflows

Cherie also joined a roundtable on what AI is changing — and not changing — for creative work.

Roundtable discussion at WaytoAGI Global AI Conference

She shared a story from inside MiniMax: a PM on the team — not a trained engineer, not a trained designer — built an experimental open-world AI interaction project end-to-end through "vibe coding" and "vibe videoing," and turned it into a subscription product generating real revenue on Bilibili. Her takeaway: as the tools get easier, the scarce ingredients are taste and a deep feel for a specific domain.

Thank you to WaytoAGI for putting together a thoughtful, well-curated day and to everyone who came up to us afterward to talk through what they're building. The conversations in Tokyo were specific, technical, and generous — exactly the kind of room we wanted to be in.