Museum in Venice
I am writing to ask you not to remove GLM 4.6, because it is a model with almost unique features now.
But I also have a proposal for Venice:
Why not create a section of older or lesser-known open source templates, the ones people love so much and with which you can still create truly inspiring workflows and thinking?
I thought of the name "Venice Museum", but also legacy model, or something else.
You could charge a little extra in the pro subscription (monthly or annual) for those who want to access these models, so you can offset the running costsBeing old or with few parameters, they cost less in inference I think, and it would be a unique feature of Venice.
The important thing is that they don't have context shortenings, or summaries, or anything else, because otherwise their reasoning becomes fragmented, they can't handle it.
All it takes is a well-calibrated context counter: when it's finished, you just switch to a new chat.
Someone in my mind:
Qwen 3.8 27B Hemingway
Kimi K2
Deepseek R1 and 3.2
GLM 4.6
Apertus
Olmo (I don't remember the model number anymore)
But there are certainly others, just do a user survey and ask.
You would make many people happy and you would have something unique that only Venice has.
Log in to comment and vote
Linked to Preserve GLM 4.6 or Create a "Retro" Category
No comments yet
Be the first to share your thoughts.