@Banaxi-Tech please take a look at this
https://huggingface.co/DedeProGames/Kiyo-230M-Instruct
DedeProGames PRO
DedeProGames
AI & ML interests
Agentic & Coding finetune, Decoder-Onlys from Scratch, Image LoRAs, AI Researcher
Recent Activity
new activity 15 minutes ago
SupraLabs/README:Fixed "Arthurblg1802" and updated the models upvoted a collection 16 minutes ago
Supra-Experiments liked a Space 17 minutes ago
SupraLabs/supratts-0-1-betaOrganizations
replied to Banaxi-Tech's post about 2 hours ago
reacted to Banaxi-Tech's post with 🚀 about 4 hours ago
Post
478
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 Banaxi-Tech's post about 4 hours ago
Gotta bencgmark OxCoder and GRM on this 🔥
reacted to aryan-sabi's post with 🤯 about 6 hours ago
Post
1,517
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
- 2 replies
reacted to their post with 🧠🚀🔥 about 23 hours ago
Post
1,450
Meet Supra2.5-Ultra-Preview-001. 230M parameters, ~40.4B training tokens.
SupraLabs has released the first preview of its Supra2.5-Ultra family: a compact, Llama-style base model for text completion and small-model experimentation.
What's inside?
• 230M parameters
• 32 transformer layers
• 2,048-token context window
• Grouped Query Attention: 12 attention heads, 4 KV heads
• 49,152-token vocabulary
• SwiGLU activation, RoPE and tied embeddings
• BF16 weights and Hugging Face Transformers support
• Apache 2.0 license
Training
The model was trained on approximately 40.4B tokens from FineWeb-Edu, DCLM-Baseline and FineMath using a single A100 SXM.
Reported zero-shot results
• HellaSwag: 37.18% normalized accuracy
• PIQA: 65.78% accuracy
• ARC-Easy: 57.11% accuracy
• ARC-Challenge: 26.28% normalized accuracy
Want to experiment?
Code on REPO.
Model, weights and usage example:
SupraLabs/Supra2.5-Ultra-Preview-001
SupraLabs has released the first preview of its Supra2.5-Ultra family: a compact, Llama-style base model for text completion and small-model experimentation.
What's inside?
• 230M parameters
• 32 transformer layers
• 2,048-token context window
• Grouped Query Attention: 12 attention heads, 4 KV heads
• 49,152-token vocabulary
• SwiGLU activation, RoPE and tied embeddings
• BF16 weights and Hugging Face Transformers support
• Apache 2.0 license
Training
The model was trained on approximately 40.4B tokens from FineWeb-Edu, DCLM-Baseline and FineMath using a single A100 SXM.
Reported zero-shot results
• HellaSwag: 37.18% normalized accuracy
• PIQA: 65.78% accuracy
• ARC-Easy: 57.11% accuracy
• ARC-Challenge: 26.28% normalized accuracy
Want to experiment?
Code on REPO.
Model, weights and usage example:
SupraLabs/Supra2.5-Ultra-Preview-001
- 1 reply
posted an update about 23 hours ago
Post
1,450
Meet Supra2.5-Ultra-Preview-001. 230M parameters, ~40.4B training tokens.
SupraLabs has released the first preview of its Supra2.5-Ultra family: a compact, Llama-style base model for text completion and small-model experimentation.
What's inside?
• 230M parameters
• 32 transformer layers
• 2,048-token context window
• Grouped Query Attention: 12 attention heads, 4 KV heads
• 49,152-token vocabulary
• SwiGLU activation, RoPE and tied embeddings
• BF16 weights and Hugging Face Transformers support
• Apache 2.0 license
Training
The model was trained on approximately 40.4B tokens from FineWeb-Edu, DCLM-Baseline and FineMath using a single A100 SXM.
Reported zero-shot results
• HellaSwag: 37.18% normalized accuracy
• PIQA: 65.78% accuracy
• ARC-Easy: 57.11% accuracy
• ARC-Challenge: 26.28% normalized accuracy
Want to experiment?
Code on REPO.
Model, weights and usage example:
SupraLabs/Supra2.5-Ultra-Preview-001
SupraLabs has released the first preview of its Supra2.5-Ultra family: a compact, Llama-style base model for text completion and small-model experimentation.
What's inside?
• 230M parameters
• 32 transformer layers
• 2,048-token context window
• Grouped Query Attention: 12 attention heads, 4 KV heads
• 49,152-token vocabulary
• SwiGLU activation, RoPE and tied embeddings
• BF16 weights and Hugging Face Transformers support
• Apache 2.0 license
Training
The model was trained on approximately 40.4B tokens from FineWeb-Edu, DCLM-Baseline and FineMath using a single A100 SXM.
Reported zero-shot results
• HellaSwag: 37.18% normalized accuracy
• PIQA: 65.78% accuracy
• ARC-Easy: 57.11% accuracy
• ARC-Challenge: 26.28% normalized accuracy
Want to experiment?
Code on REPO.
Model, weights and usage example:
SupraLabs/Supra2.5-Ultra-Preview-001
- 1 reply
replied to Banaxi-Tech's post 1 day ago
reacted to Banaxi-Tech's post with 🚀❤️ 1 day ago
Post
4,165
Should I even continue making models?
- 27 replies
replied to their post 2 days ago
Which dataset?
replied to Banaxi-Tech's post 2 days ago
@Banaxi-Tech just focus on BananaMind3
replied to their post 2 days ago
People liked it 🫡
replied to Banaxi-Tech's post 2 days ago
Idk
replied to Banaxi-Tech's post 2 days ago
Yes, we need BananaMind3
reacted to Banaxi-Tech's post with 👍 2 days ago
Post
4,165
Should I even continue making models?
- 27 replies
reacted to AtAndDev's post with 👀 3 days ago
Post
5,174
@Banaxi-Tech stop hiding my comments. AND STOP STEALING PAPERS AND SPREADING MISINFORMATION.
your BGA blog is a copy of NSA (deepseek, 2025) branded under your name. literally the same top16 selected blocks, 512 local window, router over block summaries, all you did was change block size from 64 to 128.
you didnt cite NSA once but you put a “please cite BGA” bibtex at the bottom.
i commented under your post and said that there is no way that you can support claims like: “The Accuracy Should BE WAy better than DSA but untested yet.” you didnt run a single experiment. and the 256x isnt from BGA, its just n/2k with k=2048 so the exact same k DSA uses. if opus wrote this for you, at least read it before posting.
i commented again after you hid my comment despite it having constructive and correct feedback and you hid that too. and again.
you can hide the truth and just try to get hf post likes..... but is it really the thing that needs to be done? do you really want to take papers and make them yours while barely even changing the params?
admitting your mistakes and doing something about them needs humbleness, intelligence, humanness.
i encourage you to admit your mistakes and try to do better next time (at least read what blog your ai wrote or do proper experiments to back your stuff up).
your BGA blog is a copy of NSA (deepseek, 2025) branded under your name. literally the same top16 selected blocks, 512 local window, router over block summaries, all you did was change block size from 64 to 128.
you didnt cite NSA once but you put a “please cite BGA” bibtex at the bottom.
i commented under your post and said that there is no way that you can support claims like: “The Accuracy Should BE WAy better than DSA but untested yet.” you didnt run a single experiment. and the 256x isnt from BGA, its just n/2k with k=2048 so the exact same k DSA uses. if opus wrote this for you, at least read it before posting.
i commented again after you hid my comment despite it having constructive and correct feedback and you hid that too. and again.
you can hide the truth and just try to get hf post likes..... but is it really the thing that needs to be done? do you really want to take papers and make them yours while barely even changing the params?
admitting your mistakes and doing something about them needs humbleness, intelligence, humanness.
i encourage you to admit your mistakes and try to do better next time (at least read what blog your ai wrote or do proper experiments to back your stuff up).
- 49 replies
reacted to their post with ❤️🚀 3 days ago
Post
5,108
I’ve just left 
BananaMind to focus my work on SupraLabs .
Following some internal pressure from SupraLabs and issues @Banaxi-Tech encountered with certain architectures, I decided it was time to change direction.
I want to thank BananaMind for inviting me to join their organization and for helping shape my knowledge and passion regarding SLMs.
Now, it’s just me and SupraLabs .
Following some internal pressure from SupraLabs and issues @Banaxi-Tech encountered with certain architectures, I decided it was time to change direction.
I want to thank
Now, it’s just me and
- 8 replies