
See what a CEO not having affairs does to the release rate ?
Announcing the release of zml/llmd 20261112.0. This week, we are bringing DSpine, more speculative decoding options, faster DeepSeek prefill, Apple M1 and M2 support, more control over sampling.
And of course, lots of bugfixes!
zml.ai/posts/llmd-20261112.0…
1
1
73
Corendos retweeted
and here is @zml_ai running Llama 3.2 1B *fully natively on @Windows* (CPU for now, it's a demo).
little nugget: it was entirely cross compiled from macOS using @hermeticbuild's LLVM
we're thinking of making @zml_ai windows native (nvidia, amd, intel, what else ?)
La dynamique est folle. Il se passe quelque chose.
Announcing zml/llmd-20261106.0 with support for @RicursiveAI's Ripple and @ReflectionAI_'s Beam.
Lots of QoL bugfixes and performance improvements, async scheduling in beta.
zml.ai/posts/llmd-20261106.0…
2
34
Zig is really unstoppable
Lightpanda x @zml_ai llmd: a fully local agent in Zig.
> Lightpanda for the browser
> ZML-llmd for the model server
1.7B model, one MacBook, no cloud.
Find a Pac-Man costume 👻
> Lightpanda: 1:50, 107MB, 3 results
> Chrome: 5:09, 1.8GB, no result
1
32
Who wants to frenchmaxx and run Mistral Large 4 ? 👀
Announcing zml/llmd 20261029.0 with Day 0 support for Mistral Large 4, GLM 5.2 / 5.3 and LFM 3.
zml.ai/posts/llmd-20261029.0…
36
Announcing zml/llmd 20261023.0 with 2 new models:
- Qwen/Qwen3.8-Flash-Next
- XiaomiMiMo/MiMo-V2.6
And faster deepseek-ai/DeepSeek-V4.1-Flash
zml.ai/posts/llmd-20261023.0…
On parle beaucoup des idiots utiles de l'extrême gauche, mais je trouve qu'on parle pas assez des idiots inutiles de la droite 🤔
1
85
Announcing ZML/LLMD 20261009.0, with a focus on performance zml.ai/posts/llmd-20261009.0…
Announcing ZML/LLMD 20260929.0, with DeepSeek 4.1 Flash support
zml.ai/posts/llmd-20260929.0…
Damn
That shit so easy lol
We got access to @Qualcomm's AI100 on sunday.
How fast can we make @zml_ai (and zml/llmd) run on them?
The answer is less than 24 hours.
Bringing the number of architectures that run ZML to 12:
- CPU
- NVIDIA
- AMD
- Intel
- Apple Metal
- Vulkan
- Google TPUs
- Tenstorrent
- AWS Trainium
- Moore Threads
- Furiosa NPU
- Qualcomm NPU
And still zero Python.
3
639
In close collaboration with @poolsideai, we’re happy to announce the support of Laguna S 2.1, their new open-weight model for agentic coding and long-horizon work.
It’s available today in ZML/LLMD with DFlash support. Most importantly (and as always with us), it’s available on all the supported platforms: NVIDIA, AMD, Intel, Google TPU and Apple Metal.
It's a good model, sir.
zml.ai/posts/laguna-s-2-1/
Today we're releasing Laguna S 2.1, our most capable model to date.
It's a 118B total parameter Mixture-of-Experts model with 8B activated per token, a context window of up to 1M tokens, and thinking and no-thinking modes.
Capable enough to hold its own against models many times its size. Small enough to run on a single @NVIDIAAI DGX Spark.
Laguna S 2.1 is fully open under OpenMDW-1.1, with weights available today on @huggingface
poolside.ai/blog/introducing…
Corendos retweeted
Hot French startup ZML releases free product to speed inference across lots of AI chips techcrunch.com/2026/07/08/ho…














