DeepSeek v2.5 (Full Test): Can This Open-Source LLM Beat GPT-4 in Coding?

[@AaronBlox-h2t]

Thanks for the update....I'm checking it out now and it's pretty cool...

[@Codewello]

I admit this is the first time I've ever tried it, and it's super underrated. A lot of people talk about it.

[@garchafpv]

Hahaha all the beepol. great vid and i find tour accent soothing lol idk why

[@Codewello]

i love you 🥹,I have been trying to get rid of the accent, but it just loves me so much that it won't leave

[@AbdoZaInsert]

يا عمي تحيا مصر و تحياتي لك
بجربه بعد شوي ان شاء الله يكون رهيب مثل الي بمقطعك.

[@Codewello]

حبيبى ان شاء الله يفيدك

[@internetmarketingproductre6093]

Great review !

[@Codewello]

thanks brother 🤩🤩

[@mdubbau]

I think we need to move away from using snake and life in benchmarks. Everyone tests for these and model developers know to train their models for this. We need more advanced coding tests.

[@Codewello]

any ideas i can use?

[@Maisonier]

What hardware do we need to run that DeepSeek v2.5 model?

[@Codewello]

To utilize DeepSeek-V2.5 in BF16 format for inference, 80GB*8 GPUs are required.
From Hugging Face Page it self

[@เมฆาเตียวัชรานนท์]

is it can run locally with mac studio RAM 192 GB ?

[@QorQar]

The smallest version is about 50 GB. How much RAM does it need?!

[@Codewello]

check the HF page it say :To utilize DeepSeek-V2.5 in BF16 format for inference, 80GB*8 GPUs are required.

[@QorQar]

@@Codewello q1 how many need raaaam ?

[@QorQar]

airllm?

[@franciskapsowe9183]

It is a wrapper of sonnet 3.5

[@Codewello]

Ha? Are you joking, or are you talking about the Reflection model?

[@franciskapsowe9183]

@@CodewelloSorry I confused, you are right.

[@delta-gg]

what resources would you need to run this locally with ollama? like what ram/vram?

[@Codewello]

To utilize DeepSeek-V2.5 in BF16 format for inference, 80GB*8 GPUs are required.
i got this from hf page

[@ahmedgaber3914]

Are you Egyptian? Your accent is so familiar

[@Codewello]

Yes, brother, I'm Egyptian. Are you Egyptian too?

[@thenextension9160]

Haha big chungus model.

[@Codewello]

It's a big Chungus model, but it's not the biggest Chungus out there

[@garchafpv]

The indian accent feels like nails on vhalk board to me. This is notnthat at all

[@Codewello]

Sorry that my accent bothers you so much. I'm trying to improve it. I've been trying to get rid of the accent, but it loves me so much that it just won't leave.