MiniMax touts growth and China IPO plans, releases open-weights M3 model claiming GPT-5.5 parity

MiniMax made a multi-front move: reporting fast-growing international business, preparing a Mainland China listing to supplement its Hong Kong IPO (per Bloomberg), and releasing MiniMax M3, which it bills as the first open-weights model combining three frontier capabilities, including coding and agentic use. A widely shared report claims M3 matches or beats GPT-5.5 on benchmarks at roughly 10% of the cost, intensifying the price-performance pressure that DeepSeek's 75% cut also embodies this week.

The open-weights, low-cost positioning places MiniMax squarely in the Chinese-lab cohort — alongside DeepSeek and Alibaba's Qwen — that is competing on cost and openness rather than closed frontier supremacy. The dual-listing plan signals confidence and a desire to tap domestic Chinese capital as it eyes local rivals.

The overhang is legal: MiniMax failed to dismiss a California lawsuit from Disney, Universal, and Warner Bros Discovery alleging IP theft tied to its Hailuo image and video system. The case exposes MiniMax to the same copyright-litigation risk now dogging generative-media providers globally and could complicate international expansion and any US-facing ambitions.

For the broader market, MiniMax M3's GPT-5.5-parity-at-10%-cost claim — if it holds under independent evaluation — is exactly the kind of disruption that pressures Western labs' pricing and the capital-intensive economics behind valuations like Anthropic's $965B. The caveat is that the parity claim comes largely from the company and enthusiastic coverage; independent benchmarks and the litigation outcome are the things to watch.