正在加载视频...

视频加载失败

Same prompt, 10 months later

30,875 次观看 • 1 天前 •via X (Twitter)

22 条评论

Harshith 的头像
Harshith1 天前

wow Argon is soo accurate and real

Diego | AI 🚀 - e/acc 的头像
Diego | AI 🚀 - e/acc1 天前

It did a great job here

Orion (e/acc) 的头像
Orion (e/acc)1 天前

surprised by argon man, they really need to drop a new openai model and grok is.... well, grok

Diego | AI 🚀 - e/acc 的头像
Diego | AI 🚀 - e/acc1 天前

Grok is grok hahaha

Diego | AI 🚀 - e/acc 的头像
Diego | AI 🚀 - e/acc1 天前

🕺

Błażej Kunke 🇵🇱🇪🇺 | Practical AI in Poland 的头像
Błażej Kunke 🇵🇱🇪🇺 | Practical AI in Poland1 天前

Ok that’s really impressive- sometimes we miss how much the models have improved recently

Diego | AI 🚀 - e/acc 的头像
Diego | AI 🚀 - e/acc1 天前

Opus was wild to me

Diego | AI 🚀 - e/acc 的头像
Diego | AI 🚀 - e/acc1 天前

Person Drinking Water HTML Simulation Output Lines: 402 lines of HTML (gpt-6-astra-xhigh) 1,630 lines of HTML (Gemini 4 Argon-high) 493 lines of HTML (grok-4.7-xhigh) 926 lines of HTML (claude-opus-5.5-xhigh) Line counts include embedded CSS, JavaScript, and blank lines.

Adam Holter 的头像
Adam Holter1 天前

Opus 5.5 🗿

Diego | AI 🚀 - e/acc 的头像
Diego | AI 🚀 - e/acc1 天前

🤣

Michael Lulev 的头像
Michael Lulev1 天前

Incredible that such an amazing song was created especially for this comparison. Unbelievable.

Diego | AI 🚀 - e/acc 的头像
Diego | AI 🚀 - e/acc1 天前

Thank you so much! I try to use the styles of my favorite songs, then I send the video to Gemini 4 argon to describe it and give the lyrics

Chahat Sharma 的头像
Chahat Sharma1 天前

argon at 1,630 lines is about 4x gpt's 402 for basically the same clip. do you have the line counts from the dec run too? curious if they all got longer

Diego | AI 🚀 - e/acc 的头像
Diego | AI 🚀 - e/acc1 天前

Oh yeah

Rafa 的头像
Rafa1 天前

opus tweakin' ✌️😭

Diego | AI 🚀 - e/acc 的头像
Diego | AI 🚀 - e/acc1 天前

To be further, it was the closest to a human face

Steve • Plus 的头像
Steve • Plus1 天前

Can't wait to see this in another 6 months.

Diego | AI 🚀 - e/acc 的头像
Diego | AI 🚀 - e/acc1 天前

🫡

Aditya Kamath 的头像
Aditya Kamath1 天前

Argon gigachad

T00x00T 的头像
T00x00T1 天前

The difference is striking. The ranking here is really simple: - Opus 5.5 Xhigh, with a realistic depiction of swallowing, beautifully detailed water movement in the glass, and a fairly well-controlled bokeh effect. - Gemini 4 Argon-High, also showing swallowing and water movement in the glass, but in a more conventional way, with a simple background. - GPT 6 Astra xhigh: a simple character, visually not quite polished, and basic, almost geometric water movement in the glass, nothing more. - Grok 4.7 xhigh: a very basic character and decent but basic gestures, but with a magic glass whose water never runs out, the logic isn’t there. This is the current Artificial Analysis ranking brought to life...

zer0int (it·its) 的头像
zer0int (it·its)1 天前

Now ask it to make an RT-MRI sequence, sagittal slice, of a person saying a specific sentence, lol. Because it appears your current benchmark is on the verge of "killed by saturation". Except for Grok, I guess. 😂

Resplendent Isle 的头像
Resplendent Isle1 天前

Argon soloes

相关视频