Krayon

[Krayon @oussama166948]

@ajs6888 No , I mean just the usage limits

But from when I start using it , I don't get any requests that is flagged

[Krayon @oussama166948]

What you think about GPT-6-Sol limits ?

I think its better than GPT-5.6-Sol, OpneAI improve the limits

[Krayon @oussama166948]

Tbh i think it was a scam , because they promote themselves as like cheap optionf by making the api which people will see before subscribing , but when you subscribe you find that they steal you

[Krayon @oussama166948]

Why Chinese AI labs make there APIs so cheap while there subscription usage are much worst than USA AI labs?

[Krayon @oussama166948]

The biggest problem with the Plus is that they have 5 hours limits , and one hour limit will take almost 18% of the weekly

[Krayon @oussama166948]

@1r0nM1nd who say that they dont gave more than claude and gpt? i just say that command code promote themselves as better limits , but we actually see the opposite

[Krayon @oussama166948]

I subscribe in both Command Code GOAT plan ,and in OpenCode GO plan and OpenCode look dominating here in usage.

Command code say they gave 70$ worth of usage + better Cache Hit percentage .

but this is just for marketing .

I use OpenCode heavily , I’ve consumed 3.68B tokens, almost entirely with DeepSeek V4.1 Flash, and I’ve only spent $20.

With Command Code, I used only 1.2B tokens with the same model I use in OpenCode, yet it already consumed 21% of my monthly usage.

For Budget Users i think OpenCode Go is the best subscription per value you can get

[Krayon @oussama166948]

@ALX23uz Yes i think it make better result than the previous models but we cant say its very good

i think even DeepSeek V4.1 flash make better result

i use codex , my prompt is very good . no setup issues

[Krayon @oussama166948]

GPT-6 Sol just one shooted one of the best results i have ever seen

I test it in a task to make a ship in the sea and it cocked it so hard

it even make a phone version which is my first time see a model make

but it have some down sides in the details as you can see

it complete very fast , like one of the fastest (22 minutes ) , even beating luna which make 26 minutes

[Krayon @oussama166948]

here is luna result !!

[Krayon @oussama166948]

The downsides is like there is glitch in the ship sides

when i switch from night to light the ship change little bit

when its wavy there is some waves pierces the ship

[Krayon @oussama166948]

@ning_xiao20136 @Xiaomi i use it in opencode

[Krayon @oussama166948]

Mimo V2.6 Pro and Mimo V2.6 Flash is a complete SLOP.

I try to use it in the ship test and it give me one of the baddest results i have ever seen !!

DeepSeek V4.1 Flash make better result than this like 100 times better.

This is why @Xiaomi make him cheap , because they know they make bad result .

[Krayon @oussama166948]

Opus 5.5 dropped and they are saying it beat GPT-6 Astra in a lot of benchmarks

We finally get Fable 5.1 level but with more usage

But benchmarks dont prove anything , we need to test it

[Krayon @oussama166948]

Step 5 Preview is just weak model !!

I test it in Ship in a sea test and it make one of the dumbest result i have ever seen

It take very long time (almost two hours ) without any benefits

Artificial analysis say this model is better than KIMI K3 , but in real life its the worst

Artificial analysis is not trusted anymore !

[Krayon @oussama166948]

@hakmgpt artificial analysis is not trusted anymore

[Krayon @oussama166948]

@JosephV04824396 @harry_x_coder same . GOAT

[Krayon @oussama166948]

@SPAC89 MIMO v2.6 flash look better than the pro version . hh

[Krayon @oussama166948]

@dangvanthanh what you think about GLM 5.3 Flash ? did it on pare with deepseek?

[Krayon @oussama166948]

@harry_x_coder which model you use?

[Krayon @oussama166948]

@corline0 they all have more than 95% token hit percentage , so i think the price is almost the same

[Krayon @oussama166948]

@niel_pattin wow this is worst result then me