Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-07-31 · high
MiniMax says H3 generates 2K clips at under a third of rivals' cost, and shares jumped nearly 13% on the launch.
Why now: H3 launched July 30, and MiniMax hasn't released the weights it promised within days, so no outside benchmark has checked the cost or chip claims yet.
MiniMax launched H3 on July 30, an all-modality video model that generates 15-second 2K clips with stereo sound, per the company.
MiniMax says H3 generates 2K video at under a third of what mainstream rivals charge. That's the number that matters for anyone budgeting video generation into a product.
The model takes text, image, video, and audio as input. It can also edit existing footage or transfer motion from a reference clip, instead of only generating clips from scratch.
MiniMax also designed H3 to run on several Chinese-made chips, per the launch post.
Weights are due within days of the launch, and developers can already test the model through the Vercel AI Gateway and fal. MiniMax's stock jumped nearly 13% on the announcement.
Neither the cost figure nor the chip claim comes from an outside benchmark yet. Both come from MiniMax's own post.
The cost claim is the one to watch. A third of mainstream pricing is specific and checkable, and it will either hold once outside developers get the weights, or it won't.
Each link below shares sources, entities, or timing with this story.
25,000 fake accounts. 28.8 million Claude conversations. Six weeks. And the thing they were harvesting wasn't trivia, it was software engineering and agentic reasoning. In a June 24 letter to US senators and the White House, Anthropic alleged that operators tied to Alibaba's Q...
The largest model theft accusation in AI history. Anthropic claims three Chinese companies used 24,000 fake accounts and 16 million exchanges to systematically extract Claude's capabilities — specifically targeting agentic reasoning, tool use, and coding. MiniMax reportedly pi...
I've spent the last year assuming that if I wanted real agentic coding quality, I paid for a closed model. That assumption took a hit on June 1. MiniMax shipped M3 with a new sparse-attention architecture (they call it MSA) that handles up to 1M tokens at roughly 9x prefill an...
Google added K3 to Vertex Model Garden deployable on Nvidia Blackwell-based A4 and A4X VMs, announced days after OSTP director Michael Kratsios publicly accused Moonshot of large-scale distillation from Anthropic's models. Moonshot's open-weight license requires commercial clo...
Moonshot AI released Kimi K3, a sparse mixture-of-experts activating 16 of 896 experts per token. That's about 1.8% of the pool live at any moment, with a 1M-token context window and native vision. Two new architectural pieces show up: Kimi Delta Attention and Attention Residu...
OpenAI cut GPT-5.6 Luna roughly 80%, from $1 to $0.20 per million input and $6 to $1.20 output. Anthropic priced Opus 5 at $5/$25 per million, half of Fable 5. The trigger is DeepSeek, Zhipu's GLM-5.2 and Moonshot's Kimi K3 landing 60-90% below US flagship pricing, with DoorDa...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.