DeepSeek-V3 Engineer: This New AI Coding Agent is EASY TO USE, PORTABLE & LOW-COST!

[@MrMacxter]

Thx ! And with Bolt.diy ?? An another video with deepseek v3 ?

[@maniksahdev4292]

Cannot integrate this into bolt.diy as of yet, I've tried lol

[@blenderdad]

i don't think i've ever been so early to a video. I love Deepseek because of the code quality and the price, though the designs it produces for me sometimes look a bit "old"

[@haraldwolte3745]

What are the settings its hard to see in the video. The hyperbolic url and model name etc

[@brentpope1497]

Is there any advantage using this over cline? What happens when you hit the context window max?

[@ChefBrianCooks]

not really cline is better. roo cline is better too.

[@buggi666]

Thanks!

[@AICodeKing]

Thanks for the support!

[@jacquesdupontd]

I have a "No body" error in cline when using deepseek with Hyperbolic base url, anyone else ? Thx

[@AICodeKing]

It happens when the API is down. Try a bit later.

[@jacquesdupontd]

@AICodeKing thanks

[@silviudx]

​@@AICodeKingIs it working also in roo-cline?

[@Augmented_AI]

Please ask them to make it a vscode extension

[@d.d.z.]

Another honest review ❤

[@gani2an1]

is cline not better ?

[@felixgraphx]

How to use deepseek v3 locally with ollama?

[@waytae1]

How to revert change please if the change is wrong?

[@lancemarchetti8673]

Always have a backup directory in your root folder and save the sessions. 1.html 2.html .3.html etc

[@carinebruyndoncx5331]

Not sure why you would use this over aider, since you will need some tech skill to deploy or use any code or app generated

[@syvern7197]

I'm sorry im new to these stuff but i wanted to ask, if the deepseek is opensource (free) then why are people calling it cheap, like its not a free and comes with a cheap price

[@mubashir3]

You are free to download and run it. If you have the required hardware. And that is a massive if. Most people do not. So, they pay someone else to run it for them.

[@syvern7197]

@mubashir3 interesting, and lets say i do want to download it, what is the pc requirements for doing so

[@mubashir3]

@@syvern7197 About 350-700GBs of RAM. And if you don't have a lot of CPU cores, it will run very very very slow. In others words, you could spend US $10K on a PC using state of the art components and it will still not be enough to run this model at a usable speed.

[@syvern7197]

@@mubashir3 My! that one model out of the question locally, well id it true that if you used it locally you can use it offline, also to use a model locally should we used ollama or there are other methods too

[@sm373]

Also the electricity costs of running the PC and degradation of the hardware make it not practical

[@haraldwolte3745]

Are there other providers that host deepseek model if we don't trust deepseek themselves with our data?

[@DemocracyDecoded]

Wow I cannot use gemini 2.0 and deepseek v3, they keep going down, gemini says limit reached in cline and i tried roo-cline too same problem, i tried resetting the API key, same issue, and deepseek crashes... 😂😂😂

[@thesaltyone4400]

Its the API for Gemini it's hard capped on the free at 8,000 tokens and as for deepseek if you use the API key it's not free unless you use Ollama and can run the 1tb model

[@DemocracyDecoded]

@thesaltyone4400 ya but the advertisement says free usage and people are reviewing it as its free, but really its capped like you said, which ollama model is 1tb damn that's a lot.

[@thesaltyone4400]

​@@DemocracyDecodedthat's the deepseek v3 model it is free if you can fit the 670gb download and have the hardware to run it. That's probably what King usually does.

[@thesaltyone4400]

I thought I replied a second ago but to answer your question. The Deepseek model is 670gb! If you can download it and run it on your system then it would be free. I suspect King usually downloads the models and uses them when you shows it off

[@TomHimanen]

Can we use DeepSeek v3 with Ollama locally?

[@syvern7197]

good question, i would like to hear the answer for thart too

[@cryptotrader5072]

you can but that's would cost you alot of money like thousands of dollars

[@TomHimanen]

@@cryptotrader5072 You mean because of GPU requirements? I think that there will be DeepSeek mini, which will be runnable on about 24 gigs of VRAM. Lots of open source models get scaled down dramatically by open source community.

[@syvern7197]

@@cryptotrader5072 the money is for the hardwares requirements only not the api recalling or the model itself, right?

[@1voice4all]

This one seems similar to Nexa which has a gradio UI. I need to merge FrameWise and Nexa.

[@hasanaqeelabd-alabbas3180]

I have 0 programming or knowledge about these programs you showing , i really dont know what to do

[@u007james]

please do a deepseek deep thinking review

[@GarthScaysbrook]

Awesome :)

[@chadpogs7973]

Wow!!!

[@다루루]

hmm...

[@vs-im]

Thx

[@vs-im]

Молодец

[@k4ito.cc3]

Second!

[@animationgaming8539]

🪥