Lucas van de Runstraat

[Lucas van de Runstraat @Lucas09aoht]

@ns123abc @BernieSanders What will they pause for? When will they be able to continue? If the labs can’t improve their models they should just go to another country, otherwise they will definitely go bankrupt.

[Lucas van de Runstraat @Lucas09aoht]

@LuminaBench I guess I’m unlucky..

[Lucas van de Runstraat @Lucas09aoht]

@AdamHoltererer Could be, difference in quality is really small. Maybe it needs some more testing.

[Lucas van de Runstraat @Lucas09aoht]

@LuminaBench What plan are you on? I am on pro 20x and the limits are like 250 prompts a day.

[Lucas van de Runstraat @Lucas09aoht]

@AdamHoltererer No, but also not obviously worse. They are both pretty good. What do you think of this? If it's mini thats very impressive I think. Personally, I like the mona-lisa-1 output more.

[Lucas van de Runstraat @Lucas09aoht]

@astropol0 Openai ai detector online confirms that it’s an openai model

[Lucas van de Runstraat @Lucas09aoht]

Another mona-lisa1 output

“Create a photorealistic press photograph of Steve Jobs presenting the original iPhone on stage at Macworld 2007. Accurately recreate his appearance, clothing, presentation style, stage environment and the technology appropriate to that exact event.”

[Lucas van de Runstraat @Lucas09aoht]

@R2Cdev_ I feel like grok imagine 2 ranks too high on this. It doesnt look that good compared to gpt image 2. Mona-lisa-1 will win even easier.

[Lucas van de Runstraat @Lucas09aoht]

Some more mona-lisa-1 image outputs from the arena.

[Lucas van de Runstraat @Lucas09aoht]

@TimJayas Imagine when the new image model is released

[Lucas van de Runstraat @Lucas09aoht]

@LuminaBench I think it looks pretty good, noticeable improvement.

Do you think this will be Image-2.5?

Some outputs that I got:

[Lucas van de Runstraat @Lucas09aoht]

@ai_for_success Maybe composer 3 too?

[Lucas van de Runstraat @Lucas09aoht]

@ChrisGPT - image model?
- astra?
- ultrafast mode?

[Lucas van de Runstraat @Lucas09aoht]

@bradshannon How are the limits for imagine?

[Lucas van de Runstraat @Lucas09aoht]

@haider1 Maybe the new image model too? Mona-lisa-1 has been spotted in the arena.

[Lucas van de Runstraat @Lucas09aoht]

A few more tests with mona-lisa-1 (new openai image model) in LM Arena.

What do you think?

[Lucas van de Runstraat @Lucas09aoht]

Make it look like a real award-winning Swiss/European design studio produced it.

[Lucas van de Runstraat @Lucas09aoht]

Testing what seems to be a new OpenAI image model in LM Arena, compared with the current GPT Image 2 model.

Pretty noticeable improvement in typography, graphic design and overall coherence.

What do you think?

Same prompts, prompts in the replies.

[Lucas van de Runstraat @Lucas09aoht]

schedule block. Everything must align precisely to the grid. Use only black, off-white and one fluorescent accent color. No decorative imagery except one abstract geometric form. All text must be perfectly legible, consistently typeset and intentionally kerned.