Exo Project looks to still be alive and "EXO 1.0 will be open-source soon", probably, maybe?

[Hedda]

FYI, it sounds like the "exo" AI cluster project might still be alive and will soon release EXO 1.0 according to their blog and Twitter posts:

• https://blog.exolabs.net/nvidia-dgx-spark/

Seems like they got early access to two Nvidia DGX Spark units as well as on Apple Mac Studio and and tested Exo clusters on those:

• http://x.com/exolabs/status/1978525767739883736

  • https://x.com/exolabs/status/1978525957825687877

    • https://x.com/exolabs/status/1978525955225276505

But no communication from startup founders AlexCheema or MattBeton on their GitHub repos for any exo projects:

• Exo Project Status exo-explore/exo#819

  • https://github.com/exo-explore

However Matt Beton posted a video and comment on Twitter that "EXO 1.0 will be open-source soon™" on his personal Twitter coount on the 22nd of august 2025:

• https://x.com/MattBeton/status/1958946396062851484

Alex Cheema also posted on his Twitter in the end of September this year saying they are seeking/recruiting engineers in London:

• https://x.com/alexocheema/status/1969437877198077958

Linear scaling achieved with multiple DeepSeek v3.1 instances. 4x macs = 4x throughput.

2x M3 Ultra Mac Studios = 1x DeepSeek @ 14 tok/sec

4x M3 Ultra Mac Studios = 2x DeepSeek @ 28 tok/sec

DeepSeek V3.1 is a 671B parameter model - so at its native 8-bit quantization, it requires ~700GB of memory to run the model. EXO puts half of the layers on each device, combining their memory. EXO uses MLX distributed with TB5 interconnect, optimized for Apple Silicon.
If we need higher throughput, adding two more devices lets us serve more users at once.
@exolabs
handles all of this seamlessly - adding more devices to the cluster for linear scaling as we need it.

The new EXO 1.0 will be open-source soonTM

News found via links posted here where it had seams like exo labs startup company gone into steelth mode for temporary secretiveness:

• Exo Project Status exo-explore/exo#819

Not sure if now a case of "too little and too" late as people behind company ignored community communication and pull requests, etc..?

• https://github.com/exo-explore/exo/pulls

  • Upstream contributions from exo-extended fork (ongoing PR) exo-explore/exo#852

    • Update project with downstream fixes exo-explore/exo#825

There are also forks of exo which implement those PRs as well as unique features or upstream/downstream fixes. See example:

• https://github.com/ranjithrajv/exo-extended

PS: @geerlingguy This is in reference to AI Cluster and inability to test previous versions due to it no longer being publicly maintained:

• Test Exo for clustered AI beowulf-ai-cluster#3

[geerlingguy]

I sure hope they will come out of that stealth mode, Exo looked very promising initially, but all my testing got close to working great, but never actually functioned besides tiny demos.