salah eddine bendra

[salah eddine bendra @BendraSalah]

@mapte_fireball Strata its perfect but some opttimisation i did made it faster 2/3 times only in json config .

[salah eddine bendra @BendraSalah]

@dhh its do it first perfection later , thats the chinese way of thinking

[salah eddine bendra @BendraSalah]

@aktsmm which is our idea and projecta that run into their platform we are giving them the gold of our mind and qe are paying them money thats stupidoty to accept it

[salah eddine bendra @BendraSalah]

@aktsmm with club3090 and some opt..that allow me to make freely many tests and quality is thede at 90% of the most exp llm.. this tech cant scale now and also there ia important thing the usa business are very sneaky they putting the most exp token prices to steal our ip

[salah eddine bendra @BendraSalah]

@aktsmm Welcom to the club . I have only 2 of rtx 3060 12gb 100gb ram flash next run arround 50 t/s with crx 256k .with strata but with some opt ..and have 2 of rtx 3090 and can run 3/4 coding agent in paralel with qwen 3.8 27b and 200k ctx with at least 60 t/s

[salah eddine bendra @BendraSalah]

@coldniko @yuanhao I hope.u can focus on open infrastracture like linux not mac

[salah eddine bendra @BendraSalah]

@kimmonismus its not same game USA is the past ...China is the Futur

[salah eddine bendra @BendraSalah]

@midudev its fully scams to use api from vendors

[salah eddine bendra @BendraSalah]

@coldniko @yibie Are u going to release qwen 3.8 27b for strata it will be very interesting

[salah eddine bendra @BendraSalah]

@coldniko so many false claimes today about AI , I didnt beleive STRATA until i tried it in old server z620 100gb ram and 2 of rtx 3060 12gb even one work and its fast 40 t/s thats crazy i like also the calibration ..thanks Niko its great job

[salah eddine bendra @BendraSalah]

@satyanadella Really!!!?

[salah eddine bendra @BendraSalah]

@badlogicgames Dear cant find the exact link can help

[salah eddine bendra @BendraSalah]

@TeksEdge You should test before talk , its useless model for coding

[salah eddine bendra @BendraSalah]

@hilbertspaess Well done Jacob .Dario The Evil ! Unmasked.

[Jacob Coxon @hilbertspaess]

I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

[salah eddine bendra @BendraSalah]

@karminski3 Thanks a lot of your effort i wished if you provide at the begining your result of test in engish tex .

[salah eddine bendra @BendraSalah]

@gmi_cloud @OpenRouter @MiniMax_AI The worst cloud service the api are aggresively rate limited even for simple task the worst paiment of 10 usd i made in the net .

[salah eddine bendra @BendraSalah]

@ggerganov Can we get the Ultimate version of with less than 24GB but run with full context and good speed of token/second. thats would be great ...

[salah eddine bendra @BendraSalah]

@OpenRouter @deepseek_ai The BOOM

[salah eddine bendra @BendraSalah]

@RebellioMarket https://t.co/IXRb0xn62z