Qwen 2.5 Coder - 7B (Fully Tested): Can It Crush Deepseek 2.5 & Claude Sonnet as Code assistant ?

[@AmrAbdeen]

the 32b model is now up on ollama. 14b model as well

[@Codewello]

going to review it today

[@AmrAbdeen]

@Codewello good luck

[@karim_1567]

Great Video. Greetings from Egypt ya Hosam

[@Codewello]

Thank you, Karim! I'm also from Egypt :D

[@meliodas21888]

same here xd

[@AaronBlox-h2t]

Coll stuff. Thanks.

[@Codewello]

Glad you liked it :D

[@michaelpoblete1415]

how about using qwen 2.5 instruct 32gb, not the coder variant. see how it stacks with deep seek coder and sonnet.

[@Codewello]

might do a video about it

[@bgeneto]

Should compare with DeepSeek 2.0 Lite (16B parameters I think)

[@Codewello]

Alright, I might do it, but I focus on the newest models.

[@vickeythegamer7527]

Is pretty better that as 7b model is being compared with models 20 to 30 times its size and on decent hardware it can be used as auto complete as its fast even for 7b it run supper fast and thats very important for auto complete in have an underpower rtx 4050m and i use 1.5 and 7b as my auto complete and code generator

[@Codewello]

That’s actually an excellent usage for it. The 7B model acts like a medium-sized model, though it’s still unclear how they achieved that.

[@acedusse5119]

Are all tests done on zero-shot only?

[@Codewello]

yes

[@IlllIlllIlllIlll]

Can you try qwen with aider and pear

[@Codewello]

Yes, sure! I was planning to use Pear, but Aider is a good idea

[@mrorigo]

Error at line 16, hence the code won't work..

[@Codewello]

Wow, what incredible focus you have LOL

[@DiaeBouchiki-xl3gy]

bro are they any uncesored coding model like who can create malware and some stuff like that

[@azgdoo5372]

🤨

[@Codewello]

I think the Qwin 2.5 32B is uncensored. Give it a shot! Also, try out Deep Seek 2.5; the coder is fantastic for Python. But I don’t think it will create malware in zero-shot. I think you have walked through it.

[@Codewello]

nice thumbnail bro 🤣

[@다루루]

ㅋㅋㅋㅋㅋㅋㅋㅋ