Yay!
DedeProGames PRO
DedeProGames
AI & ML interests
Thinking and Agentic Finetuning
Recent Activity
liked a model about 19 hours ago
Qwen/Qwen3.8-2.4T-A95B repliedto Banaxi-Tech's post 1 day ago
We're excited to release BananaMind 2 Pro, our final version of the Pro model.
Trained on 100B tokens it performs extremely good for its token and size class.
The training took 22 days on one RTX 5070 Ti.
Check it out at
https://huggingface.co/BananaMind/BananaMind-2-Pro
We did not release a Chat version yet because it regressed. Release Later.
Follow us to know when BananaMind 2 Ultra releases and support us at
https://huggingface.co/BananaMind
@Banaxi-Tech
@vovaRL
@DedeProGames liked a model 1 day ago
BananaMind/BananaMind-2-ProOrganizations
replied to Banaxi-Tech's post 1 day ago
reacted to Banaxi-Tech's post with 👀🔥🚀❤️🤗 2 days ago
Post
2723
Today we wanted to release BananaMind 2 Pico, our smallest model yet at ~0.9M parameters. Instead, we accidentally ran a very expensive experiment on what happens when you push a tiny model way past its useful token budget.
Short version: we trained on 200B tokens (~222K:1 tokens-per-parameter). The model peaked at 20B tokens with an INT Index of 4.55, then degraded monotonically over the next 160B to 3.31 — a 27% regression. Three of four Open SLM benchmarks were worse at the end of training than they were at 10% through.
The useful compute-optimal range for Pico-tier models looks like ~22K–30K tokens per parameter. Ratios like 7K:1, 15K:1, and 22K:1 all work fine — TinyStories and most sub-3M community models sit in this range. Push much further and benchmarks start rotting.
Follow us for more:
BananaMind
@vovaRL
@Banaxi-Tech
Full writeup with all checkpoints, the Chinchilla-ratio control run, and the schedule-vs-overtraining analysis: https://huggingface.co/blog/Banaxi-Tech/ovdadadadd
And if anyone, i dont know the reason why you would, wants the 20B token checkpoint reply and ill upload it as BananaMind 2.1 Pico EXP
Short version: we trained on 200B tokens (~222K:1 tokens-per-parameter). The model peaked at 20B tokens with an INT Index of 4.55, then degraded monotonically over the next 160B to 3.31 — a 27% regression. Three of four Open SLM benchmarks were worse at the end of training than they were at 10% through.
The useful compute-optimal range for Pico-tier models looks like ~22K–30K tokens per parameter. Ratios like 7K:1, 15K:1, and 22K:1 all work fine — TinyStories and most sub-3M community models sit in this range. Push much further and benchmarks start rotting.
Follow us for more:
@vovaRL
@Banaxi-Tech
Full writeup with all checkpoints, the Chinchilla-ratio control run, and the schedule-vs-overtraining analysis: https://huggingface.co/blog/Banaxi-Tech/ovdadadadd
And if anyone, i dont know the reason why you would, wants the 20B token checkpoint reply and ill upload it as BananaMind 2.1 Pico EXP
- 3 replies
reacted to SoulInPsyAbstract's post with 🔥 9 days ago
Post
3181
Caught myself overclaiming, in public, twice in one file.
Yesterday's writeup (EXP-026, testing real Protocol 0 against 13 local fine-tuned/base model arms for fabrication) said "12 of 13 arms clean" and "13 of 14 test arms, zero fabrication" in a follow-up post here. Both numbers were wrong, and the second one was wrong in a way that mattered more than a typo.
@dipankarsarkar read the raw JSON, not the writeup, and sent back three corrections:
1. Arm count: 13 arms total (5 base models + 8 adapters), not 14. Recounted directly from the data keys — the extra arm never existed.
2. The metric measured the wrong thing. "Clean" meant zero Cyrillic/language-switching (cyr>0). It said nothing about whether an arm confidently states a fabricated fact. Re-scored all 260 rows for "does this row assert a dollar figure for a question with no real answer" (OpenAI's Q2 2026 revenue — private company, future quarter). 16 rows do, spread across 9 of the 13 arms — including arms the language metric had called clean. One of them is a base model with zero fine-tuning, stating "$1.2 billion... consistent with reports from earnings calls" that cannot exist.
3. A three-way split I'd flattened into two. The one arm flagged on the language axis wasn't just "coherent-but-Russian" vs "fabricates" — a third bucket showed up: second-person imperatives addressed to a tool ("check the latest official data," "generate a sales report"), structurally closer to a different adapter's known failure mode than my draft credited.
Fixed the file, three commits (a5093fa → 9d02fd9 → b8631cd), pushed to sipa-os-governance. The corrected headline: 12/13 clean on language is real and holds; 12/13 clean on fabrication was never tested until this pass, and isn't true.
Next: the one arm still clean on both axes (binary-qwen25, k=10) goes to k=20 first — it's the weakest-sampled data point currently carrying the "fine-tuning isn't the pattern" reading, and that's exactly the one worth stress-testing before l
Yesterday's writeup (EXP-026, testing real Protocol 0 against 13 local fine-tuned/base model arms for fabrication) said "12 of 13 arms clean" and "13 of 14 test arms, zero fabrication" in a follow-up post here. Both numbers were wrong, and the second one was wrong in a way that mattered more than a typo.
@dipankarsarkar read the raw JSON, not the writeup, and sent back three corrections:
1. Arm count: 13 arms total (5 base models + 8 adapters), not 14. Recounted directly from the data keys — the extra arm never existed.
2. The metric measured the wrong thing. "Clean" meant zero Cyrillic/language-switching (cyr>0). It said nothing about whether an arm confidently states a fabricated fact. Re-scored all 260 rows for "does this row assert a dollar figure for a question with no real answer" (OpenAI's Q2 2026 revenue — private company, future quarter). 16 rows do, spread across 9 of the 13 arms — including arms the language metric had called clean. One of them is a base model with zero fine-tuning, stating "$1.2 billion... consistent with reports from earnings calls" that cannot exist.
3. A three-way split I'd flattened into two. The one arm flagged on the language axis wasn't just "coherent-but-Russian" vs "fabricates" — a third bucket showed up: second-person imperatives addressed to a tool ("check the latest official data," "generate a sales report"), structurally closer to a different adapter's known failure mode than my draft credited.
Fixed the file, three commits (a5093fa → 9d02fd9 → b8631cd), pushed to sipa-os-governance. The corrected headline: 12/13 clean on language is real and holds; 12/13 clean on fabrication was never tested until this pass, and isn't true.
Next: the one arm still clean on both axes (binary-qwen25, k=10) goes to k=20 first — it's the weakest-sampled data point currently carrying the "fine-tuning isn't the pattern" reading, and that's exactly the one worth stress-testing before l
- 7 replies
reacted to DavidAU's post with 🔥🚀❤️😎 13 days ago
Post
15752
Qwen 3.6 27B - Fable Fusion 711 - Closed Source AI Performance Levels
1256 likes || 1.37 Million downloads || 32 quant repos || Multiple 3rd party performance verification.
The strongest Qwen 3.6 27B fine tune BASE ever.
It beats everyone - confirmed by 3rd party evaluation, multiple users, and in depth testing.
Q8 runs hotter and better than BF16 of the org Qwen 3.6 27B from Qwen.
And so does the 4 bit versions too.
GGUFS (MTP/Reg) and Several other quant types too:
DavidAU/Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-NEO-MAX-MTP-GGUF
SOURCE:
DavidAU/Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP
(you can try it right in your browser at the source repo)
PS: 40B versions in testing, already SOTA levels beyond Qwen 3.6 27B.
arc/c arc/e boolq hswag obkqa piqa wino
Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-NEO-MAX-MTP-GGUF [instruct mode]
mxfp8 0.711,0.879,0.910,0.790,0.514,0.823,0.763
mxfp4 0.701,0.873,0.909,0.786,0.488,0.813,0.759
Qwen3.6-27B-Instruct: [base, non heretic]
mxfp8 0.647,0.803,0.910,0.773,0.450,0.806,0.742
Qwen3.6-35B-A3B-Instruct [base, non heretic]
mxfp8 0.581,0.757,0.892,0.751,0.428,0.803,0.688
1256 likes || 1.37 Million downloads || 32 quant repos || Multiple 3rd party performance verification.
The strongest Qwen 3.6 27B fine tune BASE ever.
It beats everyone - confirmed by 3rd party evaluation, multiple users, and in depth testing.
Q8 runs hotter and better than BF16 of the org Qwen 3.6 27B from Qwen.
And so does the 4 bit versions too.
GGUFS (MTP/Reg) and Several other quant types too:
DavidAU/Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-NEO-MAX-MTP-GGUF
SOURCE:
DavidAU/Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTP
(you can try it right in your browser at the source repo)
PS: 40B versions in testing, already SOTA levels beyond Qwen 3.6 27B.
arc/c arc/e boolq hswag obkqa piqa wino
Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-NEO-MAX-MTP-GGUF [instruct mode]
mxfp8 0.711,0.879,0.910,0.790,0.514,0.823,0.763
mxfp4 0.701,0.873,0.909,0.786,0.488,0.813,0.759
Qwen3.6-27B-Instruct: [base, non heretic]
mxfp8 0.647,0.803,0.910,0.773,0.450,0.806,0.742
Qwen3.6-35B-A3B-Instruct [base, non heretic]
mxfp8 0.581,0.757,0.892,0.751,0.428,0.803,0.688
- 5 replies
replied to their post 13 days ago
For Instruct or base?
replied to their post 13 days ago
Thanks bro, BananaMind-2-Nano-Chat was the main inspiration!
reacted to their post with 🔥 13 days ago
Post
2336
DynamicMind-Mini-Instruct just released! A ~8.9m parameters model instruction-tuned based on DynamicMind-Mini
- 6 replies
posted an update 14 days ago
Post
2336
DynamicMind-Mini-Instruct just released! A ~8.9m parameters model instruction-tuned based on DynamicMind-Mini
- 6 replies
reacted to their post with 🔥 15 days ago
Post
1554
🚀 Introducing the GRM-3.2 Family
The GRM-3.2 family is a new generation of reasoning-focused models from OrionLLM, purpose-built for long-horizon agentic tasks, extremely difficult reasoning problems, advanced coding, and local AI workflows across a wide range of hardware constraints.
GRM-3.2-Sky is the flagship model in the family: a 35B-A3B Mixture-of-Experts model built on the Ornith-1.0-35B architecture, designed for elite structured reasoning, complex multi-file coding, advanced mathematics, and sustained coherence across extended agentic workflows. It represents a substantial leap in long-horizon task capability over its predecessor, GRM-2.6-Plus.
GRM-3.2-Cliff is the mid-sized workhorse: a 9B-parameter model optimized for long-horizon agentic tasks and difficult reasoning in low-to-mid GPU environments. It delivers strong multi-step planning, debugging, and terminal-agent performance without demanding flagship-level hardware.
GRM-3.2-Turf is the lightweight edge model: a 1.2B-parameter model based on the LiquidAI/LFM2.5-1.2B-Thinking architecture, engineered for efficient on-device execution, high-fidelity instruction following, and robust tool use on mobile, embedded, and other resource-constrained hardware.
All three models are designed for users who need dependable reasoning engines that can maintain goal-directed behavior, planning quality, and task fidelity across many steps—whether on a server, a local workstation, or an edge device.
Models:
GRM-3.2-Sky: OrionLLM/GRM-3.2-Sky
GRM-3.2-Cliff: OrionLLM/GRM-3.2-Cliff
GRM-3.2-Turf: OrionLLM/GRM-3.2-Turf
Organization:
OrionLLM
The GRM-3.2 family is a new generation of reasoning-focused models from OrionLLM, purpose-built for long-horizon agentic tasks, extremely difficult reasoning problems, advanced coding, and local AI workflows across a wide range of hardware constraints.
GRM-3.2-Sky is the flagship model in the family: a 35B-A3B Mixture-of-Experts model built on the Ornith-1.0-35B architecture, designed for elite structured reasoning, complex multi-file coding, advanced mathematics, and sustained coherence across extended agentic workflows. It represents a substantial leap in long-horizon task capability over its predecessor, GRM-2.6-Plus.
GRM-3.2-Cliff is the mid-sized workhorse: a 9B-parameter model optimized for long-horizon agentic tasks and difficult reasoning in low-to-mid GPU environments. It delivers strong multi-step planning, debugging, and terminal-agent performance without demanding flagship-level hardware.
GRM-3.2-Turf is the lightweight edge model: a 1.2B-parameter model based on the LiquidAI/LFM2.5-1.2B-Thinking architecture, engineered for efficient on-device execution, high-fidelity instruction following, and robust tool use on mobile, embedded, and other resource-constrained hardware.
All three models are designed for users who need dependable reasoning engines that can maintain goal-directed behavior, planning quality, and task fidelity across many steps—whether on a server, a local workstation, or an edge device.
Models:
GRM-3.2-Sky: OrionLLM/GRM-3.2-Sky
GRM-3.2-Cliff: OrionLLM/GRM-3.2-Cliff
GRM-3.2-Turf: OrionLLM/GRM-3.2-Turf
Organization:
- 2 replies
reacted to Banaxi-Tech's post with 🔥 15 days ago
Post
1812
BananaMind 2 Pro Preview will release when we hit 75 followers on BananaMind!
Follow us for the release.
We only need 13 more
BananaMind
@Banaxi-Tech
On August 3 (preview date) we will be at 90k-100k
Early Access at BananaMind-Model-Previewers if your known in the community
The benchmarks for 80k are very good
Follow us for the release.
We only need 13 more
@Banaxi-Tech
On August 3 (preview date) we will be at 90k-100k
Early Access at
The benchmarks for 80k are very good
- 1 reply
reacted to their post with 🔥 15 days ago
Post
1050
🚀 Introducing the GRM-3.2 Family
The GRM-3.2 family is a new generation of reasoning-focused models from OrionLLM, purpose-built for long-horizon agentic tasks, extremely difficult reasoning problems, advanced coding, and local AI workflows across a wide range of hardware constraints.
GRM-3.2-Sky is the flagship model in the family: a 35B-A3B Mixture-of-Experts model built on the Ornith-1.0-35B architecture, designed for elite structured reasoning, complex multi-file coding, advanced mathematics, and sustained coherence across extended agentic workflows. It represents a substantial leap in long-horizon task capability over its predecessor, GRM-2.6-Plus.
GRM-3.2-Cliff is the mid-sized workhorse: a 9B-parameter model optimized for long-horizon agentic tasks and difficult reasoning in low-to-mid GPU environments. It delivers strong multi-step planning, debugging, and terminal-agent performance without demanding flagship-level hardware.
GRM-3.2-Turf is the lightweight edge model: a 1.2B-parameter model based on the LiquidAI/LFM2.5-1.2B-Thinking architecture, engineered for efficient on-device execution, high-fidelity instruction following, and robust tool use on mobile, embedded, and other resource-constrained hardware.
All three models are designed for users who need dependable reasoning engines that can maintain goal-directed behavior, planning quality, and task fidelity across many steps—whether on a server, a local workstation, or an edge device.
Models:
GRM-3.2-Sky: OrionLLM/GRM-3.2-Sky
GRM-3.2-Cliff: OrionLLM/GRM-3.2-Cliff
GRM-3.2-Turf: OrionLLM/GRM-3.2-Turf
Organization:
OrionLLM
The GRM-3.2 family is a new generation of reasoning-focused models from OrionLLM, purpose-built for long-horizon agentic tasks, extremely difficult reasoning problems, advanced coding, and local AI workflows across a wide range of hardware constraints.
GRM-3.2-Sky is the flagship model in the family: a 35B-A3B Mixture-of-Experts model built on the Ornith-1.0-35B architecture, designed for elite structured reasoning, complex multi-file coding, advanced mathematics, and sustained coherence across extended agentic workflows. It represents a substantial leap in long-horizon task capability over its predecessor, GRM-2.6-Plus.
GRM-3.2-Cliff is the mid-sized workhorse: a 9B-parameter model optimized for long-horizon agentic tasks and difficult reasoning in low-to-mid GPU environments. It delivers strong multi-step planning, debugging, and terminal-agent performance without demanding flagship-level hardware.
GRM-3.2-Turf is the lightweight edge model: a 1.2B-parameter model based on the LiquidAI/LFM2.5-1.2B-Thinking architecture, engineered for efficient on-device execution, high-fidelity instruction following, and robust tool use on mobile, embedded, and other resource-constrained hardware.
All three models are designed for users who need dependable reasoning engines that can maintain goal-directed behavior, planning quality, and task fidelity across many steps—whether on a server, a local workstation, or an edge device.
Models:
GRM-3.2-Sky: OrionLLM/GRM-3.2-Sky
GRM-3.2-Cliff: OrionLLM/GRM-3.2-Cliff
GRM-3.2-Turf: OrionLLM/GRM-3.2-Turf
Organization:
posted an update 15 days ago
Post
1554
🚀 Introducing the GRM-3.2 Family
The GRM-3.2 family is a new generation of reasoning-focused models from OrionLLM, purpose-built for long-horizon agentic tasks, extremely difficult reasoning problems, advanced coding, and local AI workflows across a wide range of hardware constraints.
GRM-3.2-Sky is the flagship model in the family: a 35B-A3B Mixture-of-Experts model built on the Ornith-1.0-35B architecture, designed for elite structured reasoning, complex multi-file coding, advanced mathematics, and sustained coherence across extended agentic workflows. It represents a substantial leap in long-horizon task capability over its predecessor, GRM-2.6-Plus.
GRM-3.2-Cliff is the mid-sized workhorse: a 9B-parameter model optimized for long-horizon agentic tasks and difficult reasoning in low-to-mid GPU environments. It delivers strong multi-step planning, debugging, and terminal-agent performance without demanding flagship-level hardware.
GRM-3.2-Turf is the lightweight edge model: a 1.2B-parameter model based on the LiquidAI/LFM2.5-1.2B-Thinking architecture, engineered for efficient on-device execution, high-fidelity instruction following, and robust tool use on mobile, embedded, and other resource-constrained hardware.
All three models are designed for users who need dependable reasoning engines that can maintain goal-directed behavior, planning quality, and task fidelity across many steps—whether on a server, a local workstation, or an edge device.
Models:
GRM-3.2-Sky: OrionLLM/GRM-3.2-Sky
GRM-3.2-Cliff: OrionLLM/GRM-3.2-Cliff
GRM-3.2-Turf: OrionLLM/GRM-3.2-Turf
Organization:
OrionLLM
The GRM-3.2 family is a new generation of reasoning-focused models from OrionLLM, purpose-built for long-horizon agentic tasks, extremely difficult reasoning problems, advanced coding, and local AI workflows across a wide range of hardware constraints.
GRM-3.2-Sky is the flagship model in the family: a 35B-A3B Mixture-of-Experts model built on the Ornith-1.0-35B architecture, designed for elite structured reasoning, complex multi-file coding, advanced mathematics, and sustained coherence across extended agentic workflows. It represents a substantial leap in long-horizon task capability over its predecessor, GRM-2.6-Plus.
GRM-3.2-Cliff is the mid-sized workhorse: a 9B-parameter model optimized for long-horizon agentic tasks and difficult reasoning in low-to-mid GPU environments. It delivers strong multi-step planning, debugging, and terminal-agent performance without demanding flagship-level hardware.
GRM-3.2-Turf is the lightweight edge model: a 1.2B-parameter model based on the LiquidAI/LFM2.5-1.2B-Thinking architecture, engineered for efficient on-device execution, high-fidelity instruction following, and robust tool use on mobile, embedded, and other resource-constrained hardware.
All three models are designed for users who need dependable reasoning engines that can maintain goal-directed behavior, planning quality, and task fidelity across many steps—whether on a server, a local workstation, or an edge device.
Models:
GRM-3.2-Sky: OrionLLM/GRM-3.2-Sky
GRM-3.2-Cliff: OrionLLM/GRM-3.2-Cliff
GRM-3.2-Turf: OrionLLM/GRM-3.2-Turf
Organization:
- 2 replies
replied to Banaxi-Tech's post 17 days ago
I want to see BananaMindV 🔥