Nick_MiniMax

Nick at MiniMax - AI, Models, Fast

[Nick_MiniMax @Nick_MiniMax]

@kevskgs @MarcBaiza @MiniMax_AI @cognition @creatine_cycle @thisistonydang @chappyasel @VictorSuOrtiz Honestly I was super impressed, you blew past everyone else!

[Enze Xie @xieenze_jr]

🚀 MiniMax H3 Super Acceleration in Sol-Engine🤩

We pushed H3 far beyond our previous 3–4× optimization regime — reaching 22.2× speedup for 5s video and 27.7× for 10s video vs. the published SGLang baseline.

On a single NVIDIA GB200:
• 5s 768p: 152.3s → 6.85s (22.2×)
• 10s 768p: 414.1s → 14.93s (27.7×)

But the more interesting part may be what this means economically.

Using MiniMax’s published H3 API price as a reference, we translate inference speed directly into production economics. Under an ideal fully utilized GB200 scenario, Sol-Super can serve about 525 five-second videos/hour, corresponding to roughly $210/hour of output value at the reference API price.

Assuming $5.50/GPU-hour, that implies a 97%+ GPU-only gross margin in the idealized model.

At full utilization, one GB200 could produce:
• 12.6K 5s videos/day
• 378K videos/month
• equivalent to 525 hours of finished video per month

For us, this is the bigger point of inference optimization: a 20×+ speedup does not just reduce latency — it can fundamentally change the unit economics, serving capacity, and viable business models of video generation.

🔗 https://t.co/dI7uBD1dSo

[Nick_MiniMax @Nick_MiniMax]

@26pablo7 @MarcBaiza @MiniMax_AI @cognition @creatine_cycle @thisistonydang @kevskgs @chappyasel @VictorSuOrtiz Had a great time!

[Nick_MiniMax @Nick_MiniMax]

@michael_chomsky @v0 @MiniMax_AI Open now

[Nick_MiniMax @Nick_MiniMax]

- lift weights with friends - chance to win your next months creatine supply - 1-shot planks w/
@v0
- Mini chest Maxxing challenge w/
@MiniMax_AI
if you like the gym and are a founder/engineer join us dm/comment for an invite 👀

[Nick_MiniMax @Nick_MiniMax]

@MiniMax_AI @truefoundry This is incredible

[Nick_MiniMax @Nick_MiniMax]

So who actually won? What prompts do you want me to test next?

[Nick_MiniMax @Nick_MiniMax]

I'm seeing a lot of interesting conversation about MiniMax H3 vs Seedance 2.5. Who wants to have a showdown?

[Nick_MiniMax @Nick_MiniMax]

Result 3

[Luma @LumaLabsAI]

You direct the look. The motion. The sound.
MiniMax H3 is now available in Luma Agents.
Generate up to 15 seconds of 2K video with native stereo sound, guided by text, image, video, and audio references.
More creative range. One continuous workflow with Luma.
Try it today → https://t.co/UA4zpCyIWl

[MiniMax (official) @MiniMax_AI]

we’ll be at AI Engineer After Dark on July 1st with @vercel , @merge_api, @FactoryAI , and a room full of people building the AI engineering stack.

our Research Lead, RL Training @olive_jy_song will be giving a lightning talk on post-training MiniMax M3 as part of a lineup featuring @browserbase , @neondatabase , @FireworksAI_HQ , @ExaAILabs , @sentry , @SurrealDB, and more.

see you after dark at SFMOMA.

[MiniMax (official) @MiniMax_AI]

M3 would never 🙂‍↔️

As a matter of fact, the weights are now open, too.

https://t.co/g4Ybfa2kWH

[Victoria Wu @victoriawu77]

10 minutes with @MiniMax_AI M3 and it built this magic website.

Hand-tracking particle playground. 👋

Wave your hand → Saturn rings spin, galaxies explode, hearts and flowers float in real time. 🪐

Minimax M3 somehow nailed complex + three.js @threejs gesture controls + particle systems in one go.

The coding ability of this open-weight model is seriously impressive. Honestly didn't expect it to handle this level of frontend interaction so well.

Live demo + prompt below 👇

[Guillermo Rauch @rauchg]

MiniMax M3 is now the leading open model on the Next.js agent evaluations (https://t.co/SnZ54XoRWV).

Right behind Opus & GPT5, but 10× cheaper (And 20× cheaper right now on ▲ AI Gateway!)

[Nick_MiniMax @Nick_MiniMax]

@CommandCodeAI Interested in offering MiniMax too?

[Nick_MiniMax @Nick_MiniMax]

@SkylerMiao7 👀👀