MiniMax Is Live in Pencil: Built for Volume

For years the AI conversation has been obsessed with the frontier. Which model is the smartest? Which one tops the leaderboard for use case X? This year, as enterprises adopted AI en masse, some new, equally pertinent questions entered that conversation: How much does your AI cost? Which models sit at the sweet spot of the performance-to-cost curve?

These are the right questions, because the frontier is quietly outrunning the work. The most intelligent models today are advancing physics, biology and higher mathematics. Deploying them to ideate CTA and headline copy is probably overkill. Using the latest and greatest for everything is not rigour. In fact it usually results in churning more money and compute than you need to.

The smart move is to match each brief to the leanest model that does the job well. Most briefs don't need a frontier model. They need a good model that's fast and cheap enough to run all day.

That's where these seven new MiniMax models step in. Spanning video, image and chat, they're built for near-frontier-grade outputs, with the speed that makes volume work viable, at a fraction of the usual costs.

What's New

Seven models, covering video, image and text generation.

• MiniMax H3 is the headline act: the newest Hailuo video model, ranked first for video editing with audio and top three for text-to-video and image-to-video on blind preference leaderboards at its release. Like its siblings it's lean and efficient, yet produces frontier-grade video. Its "omni-reference" mode supports multiple image references alongside video and audio inputs. Lock down characters, environments, props, costumes and styles with images. Control camera and motion with a video reference, and keep character voiceovers consistent with audio.
• MiniMax-Hailuo-2.3 generates video from text or from an image. Your workhorse for turning stills into motion.
• MiniMax-Hailuo-2.3-Fast does image-to-video at higher speed. Perfect for when you need to rapidly explore a variety of ideas in draft mode.
• MiniMax Image 1 is a fast, dependable image model that quickly generates images from text.
• MiniMax-M3 is the flagship chat and reasoning model, for complex copy and logic tasks.
• MiniMax-M2.7 handles everyday text work at lower cost.
• MiniMax-M2.7-highspeed is the same tier tuned for latency, built for high-volume runs.

Why MiniMax

These MiniMax models sit in the efficiency tier: the same class as the fast, low-cost lines from the major labs, the likes of Claude Haiku, GPT-5.6 Luna and Gemini's Flash series, where the models compete on price-performance optimisation rather than raw peak intelligence. This is the tier doing a lot of the real production work in 2026.

Four things stand out:

• Cost. MiniMax models run at a fraction of the price of frontier models, which changes the economics of high-volume generation.
• Token efficiency. MiniMax Sparse Attention cuts per-token compute at long context to roughly a twentieth of the previous generation at 1M tokens. Long briefs, long context, no penalty.
• Quality where it counts. On complex reasoning and logic tasks, output holds up at enterprise standard. Cheap doesn’t have to mean rough.
• Speed. Responses land fast, which keeps latency low for real-time work, facilitating creative experimentation and rapid iteration loops.

These models are a solid choice wherever you want to optimise for speed and cost.

How to Turn Them On

An admin enables MiniMax models under AI Governance → Settings, in Video Generation, Image Generation or Text Generation as appropriate. Models must be enabled before they appear in the platform.

Once enabled, the relevant MiniMax models appear wherever you normally do text, image or video generation in Pencil: Agents, the Editor, Workflows, and Feed Sheets.

Why This Matters

In 2026 model choice became an economic decision as much as a creative one. Using the most efficient model for each task unlocks more scope for exploration of the idea space. More concepts. More variants. More creativity. Experiment freely without burning a hole in the token budget, and save frontier models for where their higher cost is truly warranted.

Pencil's Open Garden philosophy means MiniMax sits within reach, alongside every other leading model, under the same governance and brand controls as everything else. You control which model to use, and when.

Seven MiniMax models are live in Pencil now.

Try them today: https://pro.trypencil.com/login