Shanghai-based MiniMax reports H1 2026 revenue up 283% YoY to ~$116.6M, above 2025's 159% YoY growth, as its M3 model lags behind rivals like Z.ai's GLM-5.2
MiniMax Group Inc.'s revenue surged 283% in the first half of 2026, even as the Chinese AI upstart trailed rivals such as Z.AI Co. in releasing top-tier large language models.
Context & Ripple Effects
MiniMax's growth curve is steepening as it scales: after reporting 2024 revenue of just $30.5M alongside Zhipu ahead of their Hong Kong listings, it posted $79M in 2025, up 159% — and the $116.6M it booked in H1 2026 alone now exceeds that full-year figure. A May filing showed it had also begun preparations for a Chinese IPO, with management citing $300M in ARR.
The tension in this print is that the acceleration is arriving while MiniMax's frontier standing slips: its M3 model trails Z.AI's GLM-5.2, a notable reversal for a company whose 2025 playbook was model-led, when it open-sourced MiniMax-M1 claiming productivity-task wins over DeepSeek. This year it has leaned on video instead — the July release of the H3 model with 2K stereo-sound clips — plus confirmed expanded investment in open-source tooling.
First-order effects
- H1 2026 revenue of ~$116.6M already surpasses MiniMax's entire 2025 revenue of $79M, giving the company a much stronger hand for the up-to-$1.9B equity-and-convertible-bond raise it was reported pursuing in July.
- Trailing Z.AI on top-tier LLMs puts pressure on the rumored M3 Pro — a 2.7T-parameter model reportedly slated for Q3 with plans to open-source it — as MiniMax's route back into the frontier conversation.
Second-order effects
- Z.AI's model lead stops being an automatic monetization lead: MiniMax is proving a Chinese AI listing can compound revenue faster than it compounds benchmarks, forcing rivals to defend pricing and enterprise accounts rather than just leaderboard position.
- If MiniMax does ship an open-source 2.7T M3 Pro from behind, it commoditizes the capability tier where it currently lags, pressuring closed-model rivals' ability to charge premiums at that performance level.
Third-order effects
- The pattern points toward Chinese AI startups being valued on commercial execution and capital-markets access rather than benchmark rank — with dual Hong Kong-plus-mainland listings and large convertible raises becoming the financing template.
- Open-source releases used defensively by non-frontier labs would erode the durability of any single lab's model lead, shifting competitive advantage toward whoever owns distribution and inference volume.
The trend: Chinese AI labs are entering a phase where revenue growth and capital raises are decoupling from frontier-model rankings, with open-sourcing serving as a defensive distribution play rather than a victory lap.