Raphi-2Code (@R2Cdev_) - Twitter Profile | Zamantika Mersobahis Locabet

Pinned Tweet

Raphi-2Code@R2Cdev_·13 Mar

GPT-5.4 Pro is a lot better at composing music than GPT-5.2 Pro.

English

0

3

142

Raphi-2Code retweeted

TestingCatalog News 🗞@testingcatalog·10h

Voice mode on Grok for X has been rolled out on the web and Android apps.

X@X

have you said thank you to Grok yet? voice mode now live on X for Android and web

English

0

8

99

7.2K

Raphi-2Code@R2Cdev_·4h

and on ARC-AGI-1, GPT-5.4 high is a lot better and cheaper than mini xhigh!

English

0

17

Raphi-2Code@R2Cdev_·4h

Don’t use mini - use GPT-5.4 with lower reasoning effort.

Chris@chatgpt21

GPT-5.4 Mini/Nano on ARC-AGI 2 GPT-5.4 Mini: - xHigh: 19%, - High: 13%, - Med: 4%, - Low: 1%, GPT-5.4 Mini is 3× cheaper per token, but used 3× more reasoning tokens, and preformed 3x worse than GPT 5.4 high

English

2

0

59

Raphi-2Code retweeted

Chris@chatgpt21·14h

GPT-5.4 Mini/Nano on ARC-AGI 2 GPT-5.4 Mini: - xHigh: 19%, - High: 13%, - Med: 4%, - Low: 1%, GPT-5.4 Mini is 3× cheaper per token, but used 3× more reasoning tokens, and preformed 3x worse than GPT 5.4 high

English

11

6

118

12.2K

Raphi-2Code retweeted

Tech Dev Notes@techdevnotes·15h

xAI has Removed grok-code-fast-1 model from the Available Models in Console

English

11

3

98

7K

Raphi-2Code retweeted

Ivan Fioravanti ᯅ@ivanfioravanti·16h

MLX Benchmark test Qwen3.5-35B-A3B-4bit MLX Benchmark Results M5 Max s M3 Ultra! The winner is M5 Max! 🥇 Details for each run in 🧵

English

6

4

57

3.4K

Raphi-2Code retweeted

Acer@AcerFur·18h

I love that GPT-5.4 Pro can write me 200 pages of a PDF in two days …and it’s only ~60% done oof

English

2

138

10.5K

Raphi-2Code retweeted

Espen JD@Snixtp·16h

I've seen very little discussion about how the new GPT-5.4 mini and nano compare to 5.3-Codex-Spark, so I did some research and created this. I focused only on official OpenAI sources, as there's a lot of misinformation out there. Hope you enjoy :)

Espen JD@Snixtp

x.com/i/article/2034…

English

0

1

3

116

Raphi-2Code retweeted

Espen JD@Snixtp·17h

There is also a Fast mode for Composer 2 which results in over 2x the speed. Faster than both GPT-5.4 Fast and Opus 4.6 Fast

English

0

1

44

Raphi-2Code retweeted

Espen JD@Snixtp·17h

Damn, jump from Composer 1.5 to 2 is really big

English

0

1

2

54

Raphi-2Code retweeted

testtm@test_tm7873·18h

Eyy almost perfect. :)))

English

0

1

4

86

Raphi-2Code retweeted

Microsoft AI@MicrosoftAI·18h

Meet MAI‑Image‑2. Built with creatives, for real creative work. Ranked #5 on @arena’s text‑to‑image leaderboard. Available now: msft.it/6014QUCBe

English

45

105

736

85.4K

Raphi-2Code retweeted

Arena.ai@arena·17h

Grok 4.20 Beta Reasoning makes @xAI a top 5 lab in Vision Arena. Scoring 1240, this model ranks #11 across all Vision models today. Congrats to the @xAI team for this milestone!

English

8

7

159

18.5K

Raphi-2Code retweeted

Lisan al Gaib@scaling01·18h

that looks pretty fucking good

English

16

7

323

47.5K

Raphi-2Code retweeted

Alex Greenland@ajrgd·1d

@chribjel @anomalyco @AnthropicAI when you see it

English

4

1

46

2.2K

Raphi-2Code retweeted

Arena.ai@arena·19h

Qwen 3.5 Max Preview has landed in top 10 for Arena Expert and top 15 for Text Arena. It shows particular strength in Math. Highlights: - #3 Math - #10 Expert - #15 Text Arena - Top 20 for Writing, Literature & Language, Life, Physical, & Social Science, Entertainment, Sports, & Media, and Medicine & Healthcare Congrats to the @Alibaba_Qwen team for this new milestone!

English

8

11

239

115.5K

Raphi-2Code retweeted

ρ:ɡeσn@pigeon__s·19h

(acording to AA-II yes i know i rely on this benchmark a lot but its like the only one we really even have) GPT-5.4 is a Gemini-3-Flash level base model being propped up by insane levels of RL