Raico

I talk to models more than people. building agents.

[Raico @raicodias]

Guys i have 4x weekly reset card for z code ( GLM 5.3 / 5.3 flash ) what do i do they are expiring in 9 days …..

[Raico @raicodias]

@Xanderwow_A @poteto ik she has seen it but she is not responding

[Raico @raicodias]

@MiaAI_lab Mia use the official harness provided by both the companies i think you might see better results

[Raico @raicodias]

@PawelHuryn Bro give GLM models a try with Z code harness they also explicitly told their models perform
Better in Z code harness

[Raico @raicodias]

@MiaAI_lab @Samking207 Yes they explicitly told to use deepseek harness to get the models
Best capabilities.. give one more try if your budget allows

[Raico @raicodias]

@Chromadera @7izzyx @dummerspast39 Guys it simple a billion dollar company ? Do you think it will let slip this mistake by any chance ? No right and these people here claiming they are right lmao what a fools @Chromadera its pointless fighting with these fools .

[Raico @raicodias]

@PawelHuryn Lub you bro ..

[Raico @raicodias]

I think you are the one that is retard here . If open ai has such models by themselves why ain’t they using it and people
Saying for distillation ! If you know that then you must know that distillation for open ai or doing is getting account ban and deepseek has a lot of training data if you are aware pf the token usage across all
Platform thats shows and tell why a mere flash model
Is ahead and second thing using a harness if you load the system
Prompts too then that will be loaded across all places and then every other model which we can use using api should also say them same if we replace the model
Id as well as the gateway …. I see you have an iq of even lower probably 15-20 iq at max . Nowadays people
How just know how to use ai spouts nonsense online thinking they will
Get away

[Raico @raicodias]

@dummerspast39 Trying to hide your lies after you are caught is more shameful please don’t do this ….

[Raico @raicodias]

@TimJayas The only reason why people will use haiku 5.5 is because they pair well with opus 5.5 as a subagent because people are forced to use it cause in claude subscriptions the ain’t giving deepseek models

[Raico @raicodias]

@PawelHuryn Deepseek 4.1 flash 1017 🙂 waiting for it

[Raico @raicodias]

A flash model better then GPT 5.6 sol ☠️☠️ i cant wait to see now what their pro models will be like ..

[DeepSeek @deepseek_ai]

🚀 DeepSeek-V4.1-Flash-1017 is here!

Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost. Now live on our website, app, and API.

🛠️ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model.

1/4

[Raico @raicodias]

@Porsche Its not your fault 🫂 … its just the other guy has the same date in mind

[Raico @raicodias]

@Haleeeemahh @StepFun_ai 6 hours to go …….. just 6

[Raico @raicodias]

@shakhzod1001 Bro i will give you 2k … twice a week so make sure on next week you gave me back my 5k which i gave you

[Raico @raicodias]

@Aditya_181105 By jailbreaking GLM 5.3 and tell it to attack the app

[Raico @raicodias]

@Haleeeemahh @notjazii you work at Devin right ! Tell us

[Raico @raicodias]

@MrTacticalX Can you share the Prompt

[Raico @raicodias]

@spectnfa I told you guys

[Raico @raicodias]

I think people are seriously underestimating where @grok is headed.

Here's why I believe @elonmusk and xAI eventually win the AI race:

1. Terafab changes the game. Most AI labs depend on someone else's chips and someone else's timeline. With Terafab, @elonmusk is building compute at a scale nobody else is even attempting. Fewer bottlenecks, faster training cycles.

2. The pace is unreal. Grok keeps shipping major model upgrades while others take months between releases. Speed compounds, and nobody moves faster than an @elonmusk company once it's locked in.

3. @bot is the most underrated launch in AI right now. It's not just a chatbot. It's close to having a full-fledged company at your fingertips: research, writing, coding, planning, execution. All for one subscription.

Most people haven't realized what that means yet. They will.

Compute + speed + real products people actually use. That's the formula @elonmusk has nailed.

So yeah maybe 1 year max it will be one sided soon…