Trust lol
🏗️ Building on HF
Boning Cui
Bc-AI
AI & ML interests
He/Him. I like LLM's and VLM's. I work with my other friends to make stuff. We are in year 7 and we are enthusiastic about AI. We are based in Australia 🇦🇺 (QLD Btw)
Recent Activity
reacted to Banaxi-Tech's post with 🚀 about 11 hours ago
KV1 updates:
We at https://huggingface.co/klein-ml have finished stage 1 of KV1 1.2B, we don't think we will need additional stages.
KV1 2.6B is near the midpoint of training, going well so far, its going to be our most agentic model yet.
We have released Klein Code Bench https://huggingface.co/datasets/klein-ml/klein-code-bench, a benchmark testing whether models can build complete projects from scratch. It contains a easy, intermediate, hard and max subset.
Here are some benchmarks of the current KV1 1.2B
It retains 75.8% of FP16 intelligence.
We are also announcing that we at https://huggingface.co/klein-ml are working on Klein 24B A2B which will be our first model featuring ternary weight quantization. reacted to Banaxi-Tech's post with 🔥 about 11 hours ago
KV1 updates:
We at https://huggingface.co/klein-ml have finished stage 1 of KV1 1.2B, we don't think we will need additional stages.
KV1 2.6B is near the midpoint of training, going well so far, its going to be our most agentic model yet.
We have released Klein Code Bench https://huggingface.co/datasets/klein-ml/klein-code-bench, a benchmark testing whether models can build complete projects from scratch. It contains a easy, intermediate, hard and max subset.
Here are some benchmarks of the current KV1 1.2B
It retains 75.8% of FP16 intelligence.
We are also announcing that we at https://huggingface.co/klein-ml are working on Klein 24B A2B which will be our first model featuring ternary weight quantization.Organizations
replied to mithulaartigala's post about 6 hours ago
reacted to Banaxi-Tech's post with 🚀🔥👀🤗👍😔 about 11 hours ago
Post
1,897
KV1 updates:
We at klein-ml have finished stage 1 of KV1 1.2B, we don't think we will need additional stages.
KV1 2.6B is near the midpoint of training, going well so far, its going to be our most agentic model yet.
We have released Klein Code Bench klein-ml/klein-code-bench, a benchmark testing whether models can build complete projects from scratch. It contains a easy, intermediate, hard and max subset.
Here are some benchmarks of the current KV1 1.2B
It retains 75.8% of FP16 intelligence.
We are also announcing that we at klein-ml are working on Klein 24B A2B which will be our first model featuring ternary weight quantization.
We at
KV1 2.6B is near the midpoint of training, going well so far, its going to be our most agentic model yet.
We have released Klein Code Bench klein-ml/klein-code-bench, a benchmark testing whether models can build complete projects from scratch. It contains a easy, intermediate, hard and max subset.
Here are some benchmarks of the current KV1 1.2B
It retains 75.8% of FP16 intelligence.
We are also announcing that we at
- 2 replies
replied to aryan-sabi's post about 12 hours ago
Their companies neural hat thingy
replied to SeaWolf-AI's post about 12 hours ago
yeah. guys please write your own responses!
reacted to AC-723's post with 🚀🔥 about 12 hours ago
Post
1,437
MixtureKit 1.1 is out 🚀 Turn the fine-tuned models you already have on the Hub into one Mixture-of-Experts model.
pip install mixturekit
Give it a base model and its fine-tunes (math, code, a language...). MixtureKit merges them, trains the routers that pick an expert for every token, and shows you what they learned:
🧩 3 composition methods behind one YAML: Branch-Train-MiX, Mixtral-style MoE, Branch-Train-Stitch
💻 Merges stream the experts tensor by tensor: they run on a laptop
🎯 Router training with a load balancing loss, optional LoRA
🔍 A routing visualizer: in the image, two Nile 4B experts split a mixed Egyptian sentence, Arabic script to one expert, Arabizi to the other
📦 Export to Mixtral / Qwen3-MoE, then GGUF for llama.cpp
🤗 mixturekit push uploads the result with a generated model card
✅ Llama 3, Qwen 2.5/3/3.5, Gemma 3/4 experts; tested on transformers 4.52 to 5.19
In the paper, a BTX model built with MixtureKit outperforms the dense baselines on multiple benchmarks.
▶️ Build, train and visualize an MoE in 15 min on a free Colab GPU: https://colab.research.google.com/github/MBZUAI-Paris/MixtureKit/blob/main/examples/build_your_own_moe.ipynb
📄 Paper: MixtureKit: A General Framework for Composing, Training, and Visualizing Mixture-of-Experts Models (2512.12121)
💻 Code: https://github.com/MBZUAI-Paris/MixtureKit
📚 Docs: https://mbzuai-paris.github.io/MixtureKit/
Work with @guokan-shang and @oe2015 at @MBZUAI-Paris 🙌
pip install mixturekit
Give it a base model and its fine-tunes (math, code, a language...). MixtureKit merges them, trains the routers that pick an expert for every token, and shows you what they learned:
🧩 3 composition methods behind one YAML: Branch-Train-MiX, Mixtral-style MoE, Branch-Train-Stitch
💻 Merges stream the experts tensor by tensor: they run on a laptop
🎯 Router training with a load balancing loss, optional LoRA
🔍 A routing visualizer: in the image, two Nile 4B experts split a mixed Egyptian sentence, Arabic script to one expert, Arabizi to the other
📦 Export to Mixtral / Qwen3-MoE, then GGUF for llama.cpp
🤗 mixturekit push uploads the result with a generated model card
✅ Llama 3, Qwen 2.5/3/3.5, Gemma 3/4 experts; tested on transformers 4.52 to 5.19
In the paper, a BTX model built with MixtureKit outperforms the dense baselines on multiple benchmarks.
▶️ Build, train and visualize an MoE in 15 min on a free Colab GPU: https://colab.research.google.com/github/MBZUAI-Paris/MixtureKit/blob/main/examples/build_your_own_moe.ipynb
📄 Paper: MixtureKit: A General Framework for Composing, Training, and Visualizing Mixture-of-Experts Models (2512.12121)
💻 Code: https://github.com/MBZUAI-Paris/MixtureKit
📚 Docs: https://mbzuai-paris.github.io/MixtureKit/
Work with @guokan-shang and @oe2015 at @MBZUAI-Paris 🙌
reacted to mithulaartigala's post with 🤯🔥 about 12 hours ago
replied to mithulaartigala's post about 12 hours ago
@mithulaartigala technically he could use the hf datacenters lol
replied to aryan-sabi's post about 13 hours ago
cool idea!
reacted to aryan-sabi's post with 👍🧠🔥🤯 about 13 hours ago
Post
3,059
tl;dr: new qwen finetune that simultaneously pushes the frontiers of small model capability and reasoning efficiency.
Hey everyone,
We are Sabi (https://sabi.com), a neuro-AI startup building a BCI + AI wearable that can do agentic actions using just your neural intents.
We publish two post-trained versions of Qwen 3.8 27B called Sabi-Noema and Sabi-Noema-MLA. These models push the capability frontier on <30B models with 2x reduction in verbosity, and are exclusively trained on open source environments.
These models have improved capabilities on coding, agentic and tool calling tasks, as well as hallucinating lesser. Sabi-Noema beats GPT 6 Luna-medium on DeepSWE, placing in between Luna 6 medium and high. It beats Luna-high on agentic and tool calling benchmarks like TB2.1, Tau-3 bench, ClawEval, and BFCL v4. Its hallucination rate is considerably lower than Luna, Qwen and other competing models.
These models are much more reasoning efficient. Sabi-Noema generates 50% fewer tokens per response on TB2.1 and 25% fewer tokens on DeepSWE, compared to base Qwen. Sabi-Noema-MLA goes even further by doing a post-hoc MLA conversion of Qwen's GQA attention. This conversion uses only a calibration dataset and no extra training. With BF16 latents Sabi-Noema-MLA needs 3.6x lower bytes per token. The MLA latents can be further compressed to FP8 with the total KV cache size on TB2.1 to 8.4x smaller with similar performance. To put this in perspective, when using Sabi-Noema-MLA at q4_0, on a 24GB GPU like 3090/4090 you can get a context size of ~534K where base Qwen at FP8 cache would only get to ~150K context.
We are excited to have the open source community play around with these models. We will be releasing more post-trains in the future focused on making <30B models more powerful.
HuggingFace:
Sabi-Noema: sabicap/sabi-noema
Sabi-Noema-MLA: sabicap/sabi-noema-mla
Hey everyone,
We are Sabi (https://sabi.com), a neuro-AI startup building a BCI + AI wearable that can do agentic actions using just your neural intents.
We publish two post-trained versions of Qwen 3.8 27B called Sabi-Noema and Sabi-Noema-MLA. These models push the capability frontier on <30B models with 2x reduction in verbosity, and are exclusively trained on open source environments.
These models have improved capabilities on coding, agentic and tool calling tasks, as well as hallucinating lesser. Sabi-Noema beats GPT 6 Luna-medium on DeepSWE, placing in between Luna 6 medium and high. It beats Luna-high on agentic and tool calling benchmarks like TB2.1, Tau-3 bench, ClawEval, and BFCL v4. Its hallucination rate is considerably lower than Luna, Qwen and other competing models.
These models are much more reasoning efficient. Sabi-Noema generates 50% fewer tokens per response on TB2.1 and 25% fewer tokens on DeepSWE, compared to base Qwen. Sabi-Noema-MLA goes even further by doing a post-hoc MLA conversion of Qwen's GQA attention. This conversion uses only a calibration dataset and no extra training. With BF16 latents Sabi-Noema-MLA needs 3.6x lower bytes per token. The MLA latents can be further compressed to FP8 with the total KV cache size on TB2.1 to 8.4x smaller with similar performance. To put this in perspective, when using Sabi-Noema-MLA at q4_0, on a 24GB GPU like 3090/4090 you can get a context size of ~534K where base Qwen at FP8 cache would only get to ~150K context.
We are excited to have the open source community play around with these models. We will be releasing more post-trains in the future focused on making <30B models more powerful.
HuggingFace:
Sabi-Noema: sabicap/sabi-noema
Sabi-Noema-MLA: sabicap/sabi-noema-mla
- 7 replies
replied to DedeProGames's post about 15 hours ago
nice model! keep up the good work