Exciting. Can't wait to see how it does.
Wayne Workman PRO
wayneworkman2012
PhysiQuanty's profile picture
webbrain-one-581620's profile picture
webbrain-one-916891's profile picture·
AI & ML interests
LLMs
Recent Activity
repliedto Banaxi-Tech's post 2 days ago
We ARE Announcing BananaMind 3 Ultra.
400M parameters, designed to beat LFM 2.5 1.2B.
This model is a usable small agent. Not a demo model like our 3M models.
FP8. KV1. Optimization. FactBase DCLM. Fable 5.2 for maximum speed kernels. TST.
This model will M.
What do you think? 50+ INT Index on the Open SLM leaderboard? BananaMind 3 is coming before christmas.
posted an update 3 days ago
I am happy to inform the HF community of new additions to the WW2Bench Leaderboard. Newest measured models:
- Qwen 3.5 27B
- Qwen 3.6 27B
- Qwen 3.8 27B
- Qwen 3.5 4B
- Qwen 3.5 2B
- KLM 1.6
- Boris 2 Instruct
- BananaMind 2 Pro
- Supra2 100M Instruct
- Claude Sonnet 5.5
- Claude Haiku 5.5
- Gemma 4 31B
The results are somewhat surprising! updated a Space 3 days ago
wayneworkman2012/ww2bench-leaderboardOrganizations
replied to Banaxi-Tech's post 2 days ago
posted an update 3 days ago
Post
90
I am happy to inform the HF community of new additions to the WW2Bench Leaderboard. Newest measured models:
- Qwen 3.5 27B
- Qwen 3.6 27B
- Qwen 3.8 27B
- Qwen 3.5 4B
- Qwen 3.5 2B
- KLM 1.6
- Boris 2 Instruct
- BananaMind 2 Pro
- Supra2 100M Instruct
- Claude Sonnet 5.5
- Claude Haiku 5.5
- Gemma 4 31B
The results are somewhat surprising!
- Qwen 3.5 27B
- Qwen 3.6 27B
- Qwen 3.8 27B
- Qwen 3.5 4B
- Qwen 3.5 2B
- KLM 1.6
- Boris 2 Instruct
- BananaMind 2 Pro
- Supra2 100M Instruct
- Claude Sonnet 5.5
- Claude Haiku 5.5
- Gemma 4 31B
The results are somewhat surprising!
replied to their post about 1 month ago
if I knew the served precision, It would save me the time of comparing the probabilities of a remote qwen 27B against my local native-precision 27B, and likewise for gemma4 31b. I am running qwen locally in native precision for the accuracy. The accuracy is what I'm after. I'd like to expand - using both my local GPU /w native precision 27B AND a remote qwen of the same precision. I don't want reduced precision for this work I'm doing.
It seems to me like they want our eyes to turn into hearts when we see "Qwen 27B" being served, but they aren't telling us they quantized it. I could run quantized qwen at twice the speed of native precision qwen, I don't want quantized.
posted an update about 1 month ago
Post
3,207
To all the inference providers behind the HuggingFace router - if you want me to use your services you will have to serve native-precision models. I'm not paying for FP8 when I can run in bf16 locally.
- 6 replies
replied to their post about 1 month ago
Can someone please explain to me what is wrong with my HF bill? Is inference free now? And if so, who can I thank?
I see you mentioned in the README using gpt and claude. Have you determined this is ok under anthropic's terms? I suppose it is because Anthropic offers no model training framework and therefore it's not competing ?? Not sure on this.
reacted to medmekk's post with ❤️ about 2 months ago
Post
4,406
🚀 Introducing Halo 1.0
Today, we are open-sourcing Halo, the training framework we use to train every model at White Circle.
It comes with:
🧠 Full post-training stack: SFT, DPO/KTO/SMPO, reward modeling, GRPO, distillation
🤖 Async multi-turn RL with vLLM/SGLang rollouts and sandboxed tool use
⚡ ~2.8× TRL throughput on 8× B300 (EP+FSDPv2, FA4, fp8/fp4)
🤗 Dense HF models + 15 MoE families (Qwen, GLM, Mistral, DeepSeek-V4…)
🛠️ One halo command, prebuilt Docker images, and docs for humans and agents
💻 https://github.com/whitecircle/halo
Try it and tell us what you're training
Today, we are open-sourcing Halo, the training framework we use to train every model at White Circle.
It comes with:
🧠 Full post-training stack: SFT, DPO/KTO/SMPO, reward modeling, GRPO, distillation
🤖 Async multi-turn RL with vLLM/SGLang rollouts and sandboxed tool use
⚡ ~2.8× TRL throughput on 8× B300 (EP+FSDPv2, FA4, fp8/fp4)
🤗 Dense HF models + 15 MoE families (Qwen, GLM, Mistral, DeepSeek-V4…)
🛠️ One halo command, prebuilt Docker images, and docs for humans and agents
💻 https://github.com/whitecircle/halo
Try it and tell us what you're training
- 1 reply
reacted to prithivMLmods's post with 👍 about 2 months ago
Post
871
VisionGuardrail EVO-2, a multimodal image-classification content-safety model based on Qwen/Qwen3.8-27B, is now available on the Hub!
Stricter image classification than before, with a dense 27-billion-parameter multimodal model, more precise reasoning, and improved captions for classifying visual media.
➠ Models: prithivMLmods/VisionGuardrail-Evo2-27B, prithivMLmods/VisionGuardrail-Evo2-27B-GGUF
➠ Collection: https://huggingface.co/collections/prithivMLmods/visionguardrail-evo2
➠ Previous Models: https://huggingface.co/collections/prithivMLmods/visionguardrail-collection
⤷ To learn more, visit the app page or the respective model pages.
Stricter image classification than before, with a dense 27-billion-parameter multimodal model, more precise reasoning, and improved captions for classifying visual media.
➠ Models: prithivMLmods/VisionGuardrail-Evo2-27B, prithivMLmods/VisionGuardrail-Evo2-27B-GGUF
➠ Collection: https://huggingface.co/collections/prithivMLmods/visionguardrail-evo2
➠ Previous Models: https://huggingface.co/collections/prithivMLmods/visionguardrail-collection
⤷ To learn more, visit the app page or the respective model pages.
I read the chat - this is really entertaining. I need to check this out.
reacted to Datdanboi25's post with ❤️ about 2 months ago
Post
3,668
THE SLM FRONTIER ADVANCES!
bench-labs/cagliostro-v3 just hit an Intelligence Index of 26.13 on the AxiomicLabs/Open_SLM_Leaderboard a 146M-param model trained completely from scratch on a single consumer GPU. That's 2nd place overall, and as far as I can tell, the most capable SLM trained on consumer hardware to date. Beating SmolLM-135m on 1/8th of the data is just silly levels of efficiency.
Big congrats to the @BenchLabs team and specifically @TobiasLogic !
bench-labs/cagliostro-v3 just hit an Intelligence Index of 26.13 on the AxiomicLabs/Open_SLM_Leaderboard a 146M-param model trained completely from scratch on a single consumer GPU. That's 2nd place overall, and as far as I can tell, the most capable SLM trained on consumer hardware to date. Beating SmolLM-135m on 1/8th of the data is just silly levels of efficiency.
Big congrats to the @BenchLabs team and specifically @TobiasLogic !
- 6 replies
replied to Datdanboi25's post about 2 months ago
Submit it to the SLM arena here: https://huggingface.co/spaces/CodeSoft/SLM-Arena
reacted to KlondikeDev's post with 👍 about 2 months ago
Post
269
Important Boris-2 news:
Boris-2 is 30B out of 200B tokens in, and it is severely behind its competitors in training.
We have determined the bug to be a configuration error. Boris-2 has been in training for ~1 week, and was projected to finish on November 3rd, 2026.
We are unfortunately going to restart training, with proper configuration.
The new projected finish date is ~15-18th of November.
We apologize for the delay.
Boris-2 is 30B out of 200B tokens in, and it is severely behind its competitors in training.
We have determined the bug to be a configuration error. Boris-2 has been in training for ~1 week, and was projected to finish on November 3rd, 2026.
We are unfortunately going to restart training, with proper configuration.
The new projected finish date is ~15-18th of November.
We apologize for the delay.
- 17 replies
posted an update 2 months ago
reacted to juiceb0xc0de's post with 🔥 2 months ago
Post
3,092
Dude, Where's My Update? I'll tell you where! ~97.6% of my BF16 parameter coordinates didn't move at all, and the ones that did overshot by ~1.33x.
It's nice to do research that doesn't end in disproving yourself once again and moving on to the next subject once in awhile.
Back to the topic, if you've ever wondered why most of your weights are basically ghosting you nearly every step when you store your weights at bf16, Dude, I Measured It.
https://huggingface.co/blog/juiceb0xc0de/intended-and-realized-updates-in-bf16-fine-tuning#dude-wheres-my-update
It's nice to do research that doesn't end in disproving yourself once again and moving on to the next subject once in awhile.
Back to the topic, if you've ever wondered why most of your weights are basically ghosting you nearly every step when you store your weights at bf16, Dude, I Measured It.
https://huggingface.co/blog/juiceb0xc0de/intended-and-realized-updates-in-bf16-fine-tuning#dude-wheres-my-update
reacted to danielhanchen's post with 🚀 2 months ago
Post
6,947
Qwen3.8-27B Unsloth GGUF is now the #1 most-liked GGUF of all time! 🤗🦥
The model hit 10M downloads and 3.7K likes in just 24 days on Hugging Face - all thanks to you.
GGUF: unsloth/Qwen3.8-27B-GGUF
Guide: https://unsloth.ai/docs/models/qwen3.8
The model hit 10M downloads and 3.7K likes in just 24 days on Hugging Face - all thanks to you.
GGUF: unsloth/Qwen3.8-27B-GGUF
Guide: https://unsloth.ai/docs/models/qwen3.8
- 5 replies
reacted to anakin87's post with ❤️🔥 2 months ago
Post
3,573
I made a 1.1M ModernBERT encoder play Doom in real time on a CPU
Some time ago, VAGO Solutions released SauerkrautLM-Doom-MultiVec-1.3M, a tiny model trained to play Doom Defend the Center scenario from 31k human gameplay examples.
My first thought: cool! I love both Doom and Small Language Models.
Then another idea: I bet I can do better :-)
What I did?
- evaluated the original model and found it's better than reported
- changed a bit the architecture
- generated SFT data with a scripted oracle
- SFT + PPO refinement on consumer hardware
Got a smaller, faster and killer model
Can even fit a floppy with int8 quantization 💾
Watch it play/read the article: anakin87/tiny-doom-defender
Some time ago, VAGO Solutions released SauerkrautLM-Doom-MultiVec-1.3M, a tiny model trained to play Doom Defend the Center scenario from 31k human gameplay examples.
My first thought: cool! I love both Doom and Small Language Models.
Then another idea: I bet I can do better :-)
What I did?
- evaluated the original model and found it's better than reported
- changed a bit the architecture
- generated SFT data with a scripted oracle
- SFT + PPO refinement on consumer hardware
Got a smaller, faster and killer model
Can even fit a floppy with int8 quantization 💾
Watch it play/read the article: anakin87/tiny-doom-defender
reacted to KlondikeDev's post with ❤️ 2 months ago
Post
2,180
The SLM Consortium has begun work on a safety dataset for Small Language Models, with the creation of the dataset being headed by @wayneworkman2012
The dataset will focus on refusals and redirects surrounding dangerous or extreme sexual content, designed to be shaped sized appropriately for SLMs, without significantly lowering benchmark performance.
More info will be out soon!
slmconsortium
The dataset will focus on refusals and redirects surrounding dangerous or extreme sexual content, designed to be shaped sized appropriately for SLMs, without significantly lowering benchmark performance.
More info will be out soon!
- 3 replies
NVIDIA has openly supported open weights models. I am interested to see how the community evolves.