DedeProGames PRO
DedeProGames
AI & ML interests
Agentic & Coding finetune, Decoder-Onlys from Scratch, Image LoRAs, AI Researcher
Recent Activity
updated a dataset about 13 hours ago
DedeProGames/lm-tetris-arena-results new activity 1 day ago
DedeProGames/kiyo-230m-chat:generate a simple JSON liked a dataset 1 day ago
DedeProGames/phone-tool-benchOrganizations
posted an update 17 days ago
Post
3,211
Hi everyone! I want to share a few updates about GRM development.
*GRM-3.4 has been canceled*
Although GRM-3.4 showed strong reasoning and coding capabilities, testing revealed several issues. It tended to overthink, and its code often contained errors that GRM-3.2-Sky handled correctly. Its tool use also needed improvement, and its terminal performance did not reach the level we were aiming for.
We repeatedly saw it rewrite files from scratch instead of using editing tools, and use shell commands to create or modify files when dedicated tools were available.
*Looking ahead to GRM-3.6*
GRM-3.6 is a possible next generation for the GRM models. We are still deciding whether it will be a single large model or a complete model family. The flagship is set to be **GRM-3.6-Moon**, a **35B-A3B** model.
*Experimental models*
Orion LLM Labs may release experimental models more regularly from now on. These experiments will help us research and test ideas for future GRM models.
Thanks for following the project!
*GRM-3.4 has been canceled*
Although GRM-3.4 showed strong reasoning and coding capabilities, testing revealed several issues. It tended to overthink, and its code often contained errors that GRM-3.2-Sky handled correctly. Its tool use also needed improvement, and its terminal performance did not reach the level we were aiming for.
We repeatedly saw it rewrite files from scratch instead of using editing tools, and use shell commands to create or modify files when dedicated tools were available.
*Looking ahead to GRM-3.6*
GRM-3.6 is a possible next generation for the GRM models. We are still deciding whether it will be a single large model or a complete model family. The flagship is set to be **GRM-3.6-Moon**, a **35B-A3B** model.
*Experimental models*
Orion LLM Labs may release experimental models more regularly from now on. These experiments will help us research and test ideas for future GRM models.
Thanks for following the project!
reacted to inflatebot's post with ๐ฅ 17 days ago
Post
3,003
i am BEGGING you people to stop shoving everything in the GGUF Hole and ship w4a8/w6a8 models for ComfyUI. Comfy has bespoke kernels for these! They're fast everywhere!
4 bits should be treated as the floor for DiT weights, if you really need 3 bits or lower you should just use a different model!
Also stop shilling nvfp4 you animals!!!
4 bits should be treated as the floor for DiT weights, if you really need 3 bits or lower you should just use a different model!
Also stop shilling nvfp4 you animals!!!
replied to Banaxi-Tech's post 22 days ago
Also what is Klein?
replied to Banaxi-Tech's post 22 days ago
@Banaxi-Tech You should get ready, because SupraLabs is cooking with Supra2.5. ๐ฅ๐ฅ๐ฅ
reacted to Banaxi-Tech's post with ๐๐ฅ 22 days ago
reacted to Banaxi-Tech's post with ๐ 23 days ago
reacted to nightmedia's post with ๐ 24 days ago
Post
2,907
The DavidAU/Qwen3.6-27B-Fable-Fusion-711
This model was assembled at NightmediaAI under the lab name:
nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451
There is ample information on the model card about the model itself.
Here are the base model metrics, as it was merged
It is a merge of different models, some abliterated, some as they were, so as a result, this source is not Heretic. DavidAU offers the source in Heretic mode here:
DavidAU/Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP
The basic idea behind this model was to build a model that can be used as a base, and has high enough IQ to merge with other models. Every model card shows details of assembly and provides a vibe check with the Nightmedia test prompts.
I merged a few samples here:
with nbeerbower/BigBubba-Qwen3.6-27B
nightmedia/Qwen3.6-27B-Fable-Fusion-711-BigBubba-717-mxfp8-mlx
with migtissera/Tess-4-27B:
nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451-Tess-mxfp8-mlx
with EpistemeAI/Reasoning-Medical-27B (metrics pending):
nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451-Medical-mxfp8-mlx
This last model was simply a blind merge, and might not be the proper formula for it, but was entertaining enough to publish as a sample.
I am populating the model card metrics as they become available.
This model was assembled at NightmediaAI under the lab name:
nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451
There is ample information on the model card about the model itself.
Here are the base model metrics, as it was merged
Qwen3.6-27B-Architect-Polaris2-Fable-B-F451
arc arc/e boolq hswag obkqa piqa wino
mxfp8 0.711,0.879,0.910,0.790,0.514,0.823,0.763
qx86-hi 0.696,0.876,0.912,0.791,0.518,0.824,0.760
qx64-hi 0.702,0.873,0.909,0.794,0.514,0.822,0.750
mxfp4 0.701,0.873,0.909,0.786,0.488,0.813,0.759It is a merge of different models, some abliterated, some as they were, so as a result, this source is not Heretic. DavidAU offers the source in Heretic mode here:
DavidAU/Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP
The basic idea behind this model was to build a model that can be used as a base, and has high enough IQ to merge with other models. Every model card shows details of assembly and provides a vibe check with the Nightmedia test prompts.
I merged a few samples here:
with nbeerbower/BigBubba-Qwen3.6-27B
nightmedia/Qwen3.6-27B-Fable-Fusion-711-BigBubba-717-mxfp8-mlx
mxfp8 0.717,0.873,0.912
mxfp4 0.707,0.877,0.912with migtissera/Tess-4-27B:
nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451-Tess-mxfp8-mlx
mxfp8 0.712,0.879,0.911,0.792,0.508,0.823,0.764
mxfp4 0.706,0.873,0.910,0.790,0.496,0.817,0.761with EpistemeAI/Reasoning-Medical-27B (metrics pending):
nightmedia/Qwen3.6-27B-Architect-Polaris2-Fable-B-F451-Medical-mxfp8-mlx
mxfp8 0.716,0.877,0.911
mxfp4 0.701This last model was simply a blind merge, and might not be the proper formula for it, but was entertaining enough to publish as a sample.
I am populating the model card metrics as they become available.
replied to their post 27 days ago
Thanks for the suggestion! I ran the CPU benchmark and Portuguese-IFEval. The F16 GGUF scored 6.54% strict prompt accuracy and decoded at 65.5 tokens/s with llama.cpp, versus 10.8 tokens/s in PyTorch FP32. The chat template passed all 9 checks. I couldnโt finish the Q4_K_M quality run, so I donโt have a valid Q4 score yet.
replied to their post 29 days ago
Ent, e dificil ahcar aqui kkkkkk
replied to their post about 1 month ago
Im from Brazil
Post
8,945
๐ Introducing Kiyo-230M-Instruct โ my compact English & Portuguese chat model!
Built on Kiyo-230M-Preview, a base model I pretrained from scratch, Kiyo now has an instruction-tuned release for everyday conversations, rewriting, and lightweight assistant experiments.
๐ง 229.7M parameters
๐ English + Portuguese
๐ฌ Native chat template for multi-turn conversations
๐ 2,048-token context window
๐ฆ 459 MB of BF16 safetensors weights
๐ Apache-2.0 license
For instruction tuning, I used 165,735 conversations from SmolTalk2, smol-smoltalk, UltraChat, and Dolci-Instruct-SFT. Training ran for two complete epochs with assistant-only loss, using a roughly 70% English / 30% Portuguese token mix.
The updated model card includes benchmarks against SmolLM2-135M-Instruct, BananaMind, and Supra2, plus example conversations, CPU measurements, training details, and a Transformers quickstart.
The results are mixed: Kiyo is competitive with these small models across the reported evaluations, while SmolLM2 and BananaMind score higher on the instruction-following benchmarks. Portuguese was a highlight in the small conversational sanity test.
This is an experimental 230M model, with plenty of room to improve reasoning, factual accuracy, and strict instruction following. Iโm excited to share it with people exploring how much we can get from small, accessible models!
๐ค Model & full README:
DedeProGames/Kiyo-230M-Instruct
๐ฌ Try the chat demo:
DedeProGames/kiyo-230m-chat
Try it in English or Portuguese โ Iโd love to see your examples, feedback, and experiments!
Built on Kiyo-230M-Preview, a base model I pretrained from scratch, Kiyo now has an instruction-tuned release for everyday conversations, rewriting, and lightweight assistant experiments.
๐ง 229.7M parameters
๐ English + Portuguese
๐ฌ Native chat template for multi-turn conversations
๐ 2,048-token context window
๐ฆ 459 MB of BF16 safetensors weights
๐ Apache-2.0 license
For instruction tuning, I used 165,735 conversations from SmolTalk2, smol-smoltalk, UltraChat, and Dolci-Instruct-SFT. Training ran for two complete epochs with assistant-only loss, using a roughly 70% English / 30% Portuguese token mix.
The updated model card includes benchmarks against SmolLM2-135M-Instruct, BananaMind, and Supra2, plus example conversations, CPU measurements, training details, and a Transformers quickstart.
The results are mixed: Kiyo is competitive with these small models across the reported evaluations, while SmolLM2 and BananaMind score higher on the instruction-following benchmarks. Portuguese was a highlight in the small conversational sanity test.
This is an experimental 230M model, with plenty of room to improve reasoning, factual accuracy, and strict instruction following. Iโm excited to share it with people exploring how much we can get from small, accessible models!
๐ค Model & full README:
DedeProGames/Kiyo-230M-Instruct
๐ฌ Try the chat demo:
DedeProGames/kiyo-230m-chat
Try it in English or Portuguese โ Iโd love to see your examples, feedback, and experiments!
posted an update about 1 month ago
Post
8,945
๐ Introducing Kiyo-230M-Instruct โ my compact English & Portuguese chat model!
Built on Kiyo-230M-Preview, a base model I pretrained from scratch, Kiyo now has an instruction-tuned release for everyday conversations, rewriting, and lightweight assistant experiments.
๐ง 229.7M parameters
๐ English + Portuguese
๐ฌ Native chat template for multi-turn conversations
๐ 2,048-token context window
๐ฆ 459 MB of BF16 safetensors weights
๐ Apache-2.0 license
For instruction tuning, I used 165,735 conversations from SmolTalk2, smol-smoltalk, UltraChat, and Dolci-Instruct-SFT. Training ran for two complete epochs with assistant-only loss, using a roughly 70% English / 30% Portuguese token mix.
The updated model card includes benchmarks against SmolLM2-135M-Instruct, BananaMind, and Supra2, plus example conversations, CPU measurements, training details, and a Transformers quickstart.
The results are mixed: Kiyo is competitive with these small models across the reported evaluations, while SmolLM2 and BananaMind score higher on the instruction-following benchmarks. Portuguese was a highlight in the small conversational sanity test.
This is an experimental 230M model, with plenty of room to improve reasoning, factual accuracy, and strict instruction following. Iโm excited to share it with people exploring how much we can get from small, accessible models!
๐ค Model & full README:
DedeProGames/Kiyo-230M-Instruct
๐ฌ Try the chat demo:
DedeProGames/kiyo-230m-chat
Try it in English or Portuguese โ Iโd love to see your examples, feedback, and experiments!
Built on Kiyo-230M-Preview, a base model I pretrained from scratch, Kiyo now has an instruction-tuned release for everyday conversations, rewriting, and lightweight assistant experiments.
๐ง 229.7M parameters
๐ English + Portuguese
๐ฌ Native chat template for multi-turn conversations
๐ 2,048-token context window
๐ฆ 459 MB of BF16 safetensors weights
๐ Apache-2.0 license
For instruction tuning, I used 165,735 conversations from SmolTalk2, smol-smoltalk, UltraChat, and Dolci-Instruct-SFT. Training ran for two complete epochs with assistant-only loss, using a roughly 70% English / 30% Portuguese token mix.
The updated model card includes benchmarks against SmolLM2-135M-Instruct, BananaMind, and Supra2, plus example conversations, CPU measurements, training details, and a Transformers quickstart.
The results are mixed: Kiyo is competitive with these small models across the reported evaluations, while SmolLM2 and BananaMind score higher on the instruction-following benchmarks. Portuguese was a highlight in the small conversational sanity test.
This is an experimental 230M model, with plenty of room to improve reasoning, factual accuracy, and strict instruction following. Iโm excited to share it with people exploring how much we can get from small, accessible models!
๐ค Model & full README:
DedeProGames/Kiyo-230M-Instruct
๐ฌ Try the chat demo:
DedeProGames/kiyo-230m-chat
Try it in English or Portuguese โ Iโd love to see your examples, feedback, and experiments!
replied to Banaxi-Tech's post about 1 month ago
@Banaxi-Tech please take a look at this
https://huggingface.co/DedeProGames/Kiyo-230M-Instruct
reacted to Banaxi-Tech's post with ๐ about 1 month ago
Post
3,821
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