Re:Clef: Open-source decision models, and new RL fine-tuning...
by bityard on Thursday October 01, 2026 @01:08PM
(#49924307)
Clef is based on Qwen3.8-27B and Clef-flash is based on Qwen3.8-9B (edit: actually Qwen3.5-9B). So, similar in spirit to Kev by my understanding, but based on a newer model.
[ Parent
| on HN ]
Re:Clef: Open-source decision models, and new RL fine-tuning...
by NitpickLawyer on Thursday October 01, 2026 @02:40PM
(#49925439)
> and Clef-flash is based on Qwen3.8-9B
There is no official qwen 3.8 9b
From the model card:
> Clef-Flash is post-trained from Qwen/Qwen3.5-9B. See Clef for the larger variant.
Re:Clef: Open-source decision models, and new RL fine-tuning...
by bityard on Thursday October 01, 2026 @03:04PM
(#49925749)
Thanks, I missed that. Fixed my comment.
Re:Clef: Open-source decision models, and new RL fine-tuning...
by ddarolfi on Thursday October 01, 2026 @02:48PM
(#49925539)
It's based on Qwen3.5-9B, maybe a typo
Re:Clef: Open-source decision models, and new RL fine-tuning...
by verdverm on Thursday October 01, 2026 @06:59PM
(#49928047)
I've been guilty of the same wishful projection
Re:Clef: Open-source decision models, and new RL fine-tuning...
by okpatil on Thursday October 01, 2026 @03:06PM
(#49925771)
Atom is 60M Param (around 133x to 400x smaller).
16ms latency. And locally run.
Why go big when you can go small ?
Re:Clef: Open-source decision models, and new RL fine-tuning...
by kamranjon on Thursday October 01, 2026 @03:46PM
(#49926218)
Cause it's not open?
Re:Clef: Open-source decision models, and new RL fine-tuning...
by okpatil on Thursday October 01, 2026 @03:51PM
(#49926270)
Good point.
To counter, most of the AI is not open. So is none of Microsoft Products. As long as they work, we keep using them.
Re:Clef: Open-source decision models, and new RL fine-tuning...
by verdverm on Thursday October 01, 2026 @07:00PM
(#49928062)
counter point, I've stopped using all closed models and harnesses as a life choice
Ai is too important and transformational to let Big Ai dominate in a closed ecosystem, thankfully the Chinese have a different mindset and approach
Re:Clef: Open-source decision models, and new RL fine-tuning...
by okpatil on Friday October 02, 2026 @12:10PM
(#49935176)
That's an absolutely fair way. More power to you !
Re:Clef: Open-source decision models, and new RL fine-tuning...
by ricardobeat on Friday October 02, 2026 @05:13AM
(#49931457)
Because with a tiny model you're skipping all the intelligence and world knowledge that makes it useful without fine tuning. `typed-decisions` is almost entirely text classification tasks.