Will Lin

PhD Student @ Hao AI Lab, Working on FastVideo

[Zico @zicojzc]

Made a realtime Draw & Guess game with FastH3.

AI joins as a player, it sees the sketch, makes a guess, then uses FastH3 to generate a matching video as proof. AI only wins if both the guess and the video are right. Played it with my friends, it got pretty funny.

Built with @AgoraIO RTC + Interactive Whiteboard and @reactorworld’s FastH3.

I’ll publish the demo and open-source it soon

[Alberto @taiuti]

Oh my god

[Susan Zhang @suchenzang]

*taps sign*

there's no pausing this gravy train, the cat is far too out of the bag

and boy was i naive to think that "enforceable standards" would be anything but regulatory capture

silly me, no time like now to update these poor priors!

[Junlin Chen @Chen94751623484]

Don't forget to check our training log
https://t.co/16VuErkx35

[vLLM @vllm_project]

K2-Horizon has day-0 support in vLLM, and IFM released intermediate checkpoints, detailed data-construction recipes, the training code, and fine-grained logs alongside the weights. 📷

512K context and Apache-2.0 from 3.7B up, with reasoning and tool calling in the checkpoints for agents and long documents.

vLLM runs the whole family, from local to large-scale data centers. Prototype on the small end, scale up without changing the stack.

Thanks @IFM_AI for opening all six and every stage that led to them, not just the last checkpoint! 📷

📷 https://t.co/JgteusHkLp

[Will Lin @wlsaidhi]

Open source 🚀🚀🚀

[Institute of Foundation Models @IFM_AI]

Introducing K2 Horizon: a connected fleet of six foundation models ranging from 0.9 billion to 375 billion parameters.

- Frontier performance: Across coding and agentic tasks, K2 Horizon delivers top-tier performance in every size class—with the 0.9B, 3.7B and 7B models setting new state of the art at their respective scales.
- Radical openness: K2 Horizon represents the largest fully open-source model launch in AI history. The fully open code, training data and recipes are a significant step forward in transparency.

Launch page: https://t.co/gg0k803SbL
Tech blog: https://t.co/g35L5xMGdS
Hugging Face: https://t.co/3Lb28JhyG9

[Will Lin @wlsaidhi]

Amazing speed by the @reactorworld team to bring online FastVideo's FastH3 API with such a cheap price 🚀

Try it out!

[reactor @reactorworld]

Introducing MiniMax FastH3 on Reactor - the latest open-source model by @MiniMax_AI.

Stream 720p with audio at $0.007/second with first frame image support, faster and cheaper than ever.

Get started now by cloning our demo app: https://t.co/7qeQBgVZ2a

[CALMDUST aka ちゃんこあ @core_tan]

MiniMax H3 WebUI ver1.5 Final
FastVideo VSA 4step ドラフトモード
1088x1920（フルHD）5秒 77.4s

色々あって描画エンジンを作り直したので、更に高画質になったよ

[MiniMax (official) @MiniMax_AI]

Fast H3 on a local device! Good news for the localhost community: open, fast, high quality video generation running on local hardware! 🥳

There are so many good recipes for running local H3 - which one should I choose? 🫨

[Hao AI Lab @haoailab]

(1/7) Last week FastVideo released FastH3. Today we bring it local: @NVIDIA DGX Spark 💚 and @Apple Silicon 🍎through MLX. 8x speedup!

FastH3 now supported on NVIDIA DGX Spark and Apple Silicon! over base @MiniMax_AI H3 for local generation!

This 15 seconds 1344×768 clip was generated on two NVIDIA DGX Sparks. Fully local. @NVIDIARTXSpark @NVIDIAAI

[Aryan @aryan_xv]

This video was generated entirely on my Mac Studio.

For this FastH3 release I worked across the local inference stack: Apple Silicon/MLX, DGX Spark, multi-Spark execution, and the cookbook/serving path.

The goal was straightforward: make it run on consumer hardware.

[Will Lin @wlsaidhi]

Amazing effort by the entire FastVideo team! Super proud of how fast we are shipping! Open source must win!

More coming soon, RTX support, FastH3 V2, and Ref2VA support! 🚀

Special shout out to @nuvalab, and @NVIDIAAI FastGen and Enterprise Products Team for the joint effort!

[Hao AI Lab @haoailab]

(1/7) Last week FastVideo released FastH3. Today we bring it local: @NVIDIA DGX Spark 💚 and @Apple Silicon 🍎through MLX. 8x speedup!

FastH3 now supported on NVIDIA DGX Spark and Apple Silicon! over base @MiniMax_AI H3 for local generation!

This 15 seconds 1344×768 clip was generated on two NVIDIA DGX Sparks. Fully local. @NVIDIARTXSpark @NVIDIAAI

[Will Lin @wlsaidhi]

Checkout FastVideo’s FastH3 dense served on @vllm_project !

Love to see our sparse distilled @MiniMax_AI H3 supported and used by every in the open source community!

More updates coming soon!

[vLLM @vllm_project]

🎬Video generation faster than playback!

🚀MiniMax H3 on vLLM-Omni + FastVideo's FastH3: a complete 10.1s MP4 - video AND synchronized audio - rendered in 8.7s!⚡️

Thanks to @MiniMax_AI for the great Minimax H3 release, the FastVideo team @haoailab for open-sourcing FastH3 and helping on the serving integration, and @NVIDIAAI for the continued sponsorship and joint optimization efforts!

[vLLM @vllm_project]

🎬Video generation faster than playback!

🚀MiniMax H3 on vLLM-Omni + FastVideo's FastH3: a complete 10.1s MP4 - video AND synchronized audio - rendered in 8.7s!⚡️

Thanks to @MiniMax_AI for the great Minimax H3 release, the FastVideo team @haoailab for open-sourcing FastH3 and helping on the serving integration, and @NVIDIAAI for the continued sponsorship and joint optimization efforts!

[Will Lin @wlsaidhi]

This doesn’t cost 6k+ a day btw :)

stay tuned

[@levelsio @levelsio]

I'm starting to get requests to sponsor this with video ads in between the slop

If you're interested let me know because it'd be nice to actually make this profitable!

We have to cover about $200,000/mo in costs

That'd be do-able with a few sponsors like:
- 20x $10K/mo
- 8x $25K/mo
- 4x $50K/mo
- 2x $100K/mo

[Naval @naval]

The people who want to censor (code, speech, money) are the bad guys.

[Lanxiang Hu @Lanxiang_Hu]

Happy to share I have been selected for the 2026 QIF. Look forward to building with the team!

[Will Lin @wlsaidhi]

@Lanxiang_Hu Congrats!

[reactor @reactorworld]

Reactor TV is here.

Check it out: https://t.co/hnCOh1ZAtg

[reactor @reactorworld]

we're back online here: https://t.co/3JZD6UpCrG