Repository navigation
Is zen/big-pickle glm 4.6? #4276
Description
Activity
Yes it is
Reacted by Roman, rqdmap, BestCodes, Ahmed Mahmoud, bon3less, Gwaboo, weakish, Maikel Frias Mosquea, 椰格, de-flandres and 9 moreill answer any other questions u have in here too
Reacted by Hamza, BestCodes, Eliah Rusin, Roman, rqdmap, Phillip Samuel K., Antar Basil, Christopher Pereira, Edgars Gogolevs, giszwk and 10 moreReacted by Hugeill answer any other questions u have in here too
I have 2 more questions if it is ok to ask them, what is the max output of the model? for how long big-pickle is gonna be free on zen?
Thank you.
uh im actually not sure the max output off the dome but it should be on models.dev
we are hoping to keep it free in perpetuity, we have done the math and it seems possible
Reacted by GeekLad, Estevão de Abreu Machado, Daniel Maricic, Chris Irineo, ak127a, Anomatomato, bailey cosier, thoughton, Krisspy Wang, Jonathan Daniels and 48 moremax output is 128k
Reacted by Hamza, BestCodes, Daniel Maricic, Chris Irineo, ak127a, bailey cosier, Phillip Samuel K., Roman, Mikec78660, Nexulo and 12 morewe are hoping to keep it free in perpetuity, we have done the math and it seems possible
Sounds awesome, thank you.
Just a curious question, I've seen glm 4.6 being offered free on some other platforms why exactly they offer it for free? What makes this model so special? I signed up for zen and GLM 4.6 is $0.60 / $2.20, this is the point that I couldn't understand tbh 😄
its just one of the better OSS models, so since we can host it ourselves we have a lot of levers to allow cheaper pricing
Reacted by christian, Anomatomato, 魔王少年(maou shonen), Jonathan Daniels, Phillip Samuel K., BestCodes, Roman, Mike, Michael New, Nexulo and 21 moreif a new oss model comes out thats way better wed prolly switch to it
Reacted by Benjamin Becker and Ricky BrentNo doubt that opencode is the best AI coding agent for the terminal, having a team like this behind it.
I don't see myself using any terminal agent other than opencode, and when I need to tap into paid models, I'll absolutlely use zen. Wish you guys all the success in what you are doing.
Reacted by maple, christian, Anomatomato, Sergio Rael, Phillip Samuel K., BestCodes, 4wl2d, Roman, Ben Lyo, Michael New and 26 moreWell thank u we appreciate it!
Reacted by Luis Marques, Saeedeeas123, Mubashir Jamali, ShinaZin, OopsRmRf, gleb, roxen, Simanta Das, Akhile E. Daniel and Elad LeviAny reasn you call it big pickle and not by its normal name?
Tbh i have no idea haha
Not an issue or a question but I want to give a big thank you and congrats on such a great product! Love you guys ❤️ !
Reacted by christian, Chris Tunbridge, Jordan Hindo, Edgars Gogolevs, Skullthought, Benjamin Becker, xelemorf, Simanta Das and Akhile E. Danielthank u!
7 remaining items
@rekram1-node has big pickle been updated to glm-4.7 ?
Reacted by themajashurka and alMubarmijReacted by alMubarmijReacted by Hugeif a new oss model comes out thats way better wed prolly switch to it
@rekram1-node has big pickle switched to DeepSeek v4 flash or glm 5.1?
Reacted by alMubarmijif a new oss model comes out thats way better wed prolly switch to it
@rekram1-node has big pickle switched to DeepSeek v4 flash or glm 5.1?
I think deepseek because I used it though the openai endpoint and it showed an error saying deepseek provider returned an error
Reacted by HugeI think deepseek because I used it though the openai endpoint and it showed an error saying deepseek provider returned an error
@inikishev i checked the tokenizer and few prompt tricks and it indeed is deepseek (i assume flash, i doubt pro)
I need to say thank you to the devs of BigPickel. I just compleated my Panic aplication for mobile and server. I did see he got confused sometimes between local machien and server.
Reacted by maple, Pourya Aliannejadi, de-flandres and HugeI think deepseek because I used it though the openai endpoint and it showed an error saying deepseek provider returned an error
@inikishev i checked the tokenizer and few prompt tricks and it indeed is deepseek (i assume flash, i doubt pro)
definitely some sort of deepseek model
Reacted by HugeThis thread heavily reminds me of the ancient parable of the blind men and the elephant.
Six blind men touch a different part of an elephant, the side, the tusk, the trunk, the leg, and each passionately argues that the beast is a wall, a spear, a snake, or a tree. Each is entirely correct about their local observation, yet fundamentally wrong about the reality.
Reacted by Daniel Maricic, Jason He and Sahib Bhaibig-pickle is changing the underlaying model, as of time of writing is
deepseek-v4-flashThey still haven't updated open code to reflect the actual context limit tho, cos it thinks 200k is the Max but it's actually around 1M
@rekram1-node Hey big pickel underline model is changing?
What is now ?@rekram1-node ok. but can you share the exact model is used when Big Pickle is selected so I could run online the exact same version (then experiment with different quants) ?
currently Big Pickle is a stealth model for Ox Alpha
@inikishev how did you know ?
Reacted by HugeNo!
Big-Pickle is just a router to a new stealth or test models.
It is like: "openrouter/free" which is variable for free models.@inikishev how did you know ?
I think that the guardrails had a bug because when I asked it, it just told me that it was Ox Alpha
Question
I noticed that the context window is 200k similar to glm 4.6, it also act and behave in a similar way.