VEHICOULE

Sup

[VEHICOULE @NeKr0w_]

@Shwepsik21 @EvanOtero Not how it works
Even 3.7 is better than glm 5.3 flash and much faster
And cost means nothing if one have a high value subscribtion, when glm sub offers not much usage compared to api on 5.3 flash, while i can use Gemini flash all day and hardly reach usage limit

[VEHICOULE @NeKr0w_]

@ChwiyaDeTout @ZixuanLi_ You have to switch https://t.co/qu3C4GNlsE provider to start plan

[VEHICOULE @NeKr0w_]

@theo Gemini flash in DSH rocks ngl, i use it with my sub throught opencodex

[VEHICOULE @NeKr0w_]

@t1llmann @dhh What mac do you have, everything up to M3 should work without any major issue, and it works like a dual boot between macos and linux

[VEHICOULE @NeKr0w_]

@jtdavies Ox alpha ?

[VEHICOULE @NeKr0w_]

@jezell Why gpui when gtk is hands down the best framework on desktop

[VEHICOULE @NeKr0w_]

@ThomasBekkers @DavidOndrej1 It's not the smartest model in one shot but excellent on retries and never get stuck
Benchmarks dont show the full story

[VEHICOULE @NeKr0w_]

@1kartikkabadi1 600h+ on my end with a goal

[VEHICOULE @NeKr0w_]

@DavidNkaha @maria_rcks No ox supports video, probably mistral, glm, mimo or even ssi

[VEHICOULE @NeKr0w_]

@aisearchio Mistral

[VEHICOULE @NeKr0w_]

@teortaxesTex I want to believe it's mistral

[VEHICOULE @NeKr0w_]

@davis7 Whats the probability it is mistral ? They used glm architecture in the past

[VEHICOULE @NeKr0w_]

@teortaxesTex Stop calling it glm model, this thing support video as well

[VEHICOULE @NeKr0w_]

@thdxr An1 is bound to die on day or another, if you choose blue at least your are not the one responsable for it

[VEHICOULE @NeKr0w_]

@John__Micheal @danieltvela The model itself is made for self hosting anyway, no one will pay for a 27b cloud model

[VEHICOULE @NeKr0w_]

@ENY66n @eliebakouch Probably DSH

[VEHICOULE @NeKr0w_]

@gianwirth DS ib hermes gives better results than any harness i've tried, and et faster iteration speed and less token
+ the ability of hermes to create skills on the fly based on your workflow is kinda sexy

[Leon @LeonWgmii]

@Telegraphprotoc signed up.

[VEHICOULE @NeKr0w_]

@PirAhmedShah_ @chribjel Not true opencode go follows deepseek pricing, i got 1b tokens for about 6€, which is like 1/10 of the monthly limit of 60€
With official deepseek api it would have really cost 5€

[VEHICOULE @NeKr0w_]

@joshhfm @Im_IrushiK High flyee is not a startup and they operate at 6x profit margin for api users