Lumina's banner
Lumina's profile picture

Lumina

@LuminaBench • 16,040 subscribers

AI News and Benchmarks, Join Our Discord for early leaks (https://t.co/t27HFvo0r9), For Tips/Contact ([email protected])

Shorts

🚨 Gemini 4 Pro vs GPT-6 Astra Pro (Pelican Riding a Bicycle SVG) Both look quite similar, maybe Astra is slightly better imo But remember this is still a Gemini 4 Pro checkpoint (argon-d), so it’ll probably improve from here Very bullish on Google rn after testing this out

🚨 Gemini 4 Pro vs GPT-6 Astra Pro (Pelican Riding a Bicycle SVG) Both look quite similar, maybe Astra is slightly better imo But remember this is still a Gemini 4 Pro checkpoint (argon-d), so it’ll probably improve from here Very bullish on Google rn after testing this out

111,136 views

🚨 GPT-6.1 Sol vs Gemini 4 Pro vs GPT-6 Astra 3D Colosseum, all on the same prompt, Gemini 4 pro is on checkpoint 3 here too Astra and the new Sol feel very similar imo

🚨 GPT-6.1 Sol vs Gemini 4 Pro vs GPT-6 Astra 3D Colosseum, all on the same prompt, Gemini 4 pro is on checkpoint 3 here too Astra and the new Sol feel very similar imo

70,307 views

🚨A new Gemini 4 Pro checkpoint might be in arena "Gemini 3.8 flash" is acting suspiciously good again I’ll show a comparison against the real Gemini 3.8 Flash shortly, but this took under 5 minutes to make

🚨A new Gemini 4 Pro checkpoint might be in arena "Gemini 3.8 flash" is acting suspiciously good again I’ll show a comparison against the real Gemini 3.8 Flash shortly, but this took under 5 minutes to make

57,527 views

🚨 Gemini 4 Pro (checkpoint 1 vs checkpoint 2) Exact same prompt "SVG of a pelican riding a bicycle" Very different styles, I prefer checkpoint 2 myself

🚨 Gemini 4 Pro (checkpoint 1 vs checkpoint 2) Exact same prompt "SVG of a pelican riding a bicycle" Very different styles, I prefer checkpoint 2 myself

51,044 views

🚨 Sonnet 5.5 vs Gemini 4 Pro vs GPT 6 Sol & Luna Same voxel pagoda prompt across all 4 OpenAI models seem to have a drastically different style though, also Gemini 4 pro is from checkpoint 2 when it was in arena a couple days ago

🚨 Sonnet 5.5 vs Gemini 4 Pro vs GPT 6 Sol & Luna Same voxel pagoda prompt across all 4 OpenAI models seem to have a drastically different style though, also Gemini 4 pro is from checkpoint 2 when it was in arena a couple days ago

43,098 views

🚨 Gemini 4 Pro (Pelican riding Bicycle SVG) This is very impressive and only took around 6 minutes to make in total The new Gemini 4 Pro checkpoint is really impressive

🚨 Gemini 4 Pro (Pelican riding Bicycle SVG) This is very impressive and only took around 6 minutes to make in total The new Gemini 4 Pro checkpoint is really impressive

42,295 views

🚨Gemini 3.6 Flash is trash I tested it on a 3D Golden Gate Bridge, and the results were awful. • I had to re-prompt it three times because it repeatedly ignored the instructions. • First attempt, instead of creating the requested .html file, it first tried to build the experience inside the Gemini app using simulations. • Then second attempt it started placing images from the web into the chat rather than actually producing the file. • Even after getting it to complete the task, the final output was dramatically worse than Gemini 3.1 Pro, which is 5 months old and now not even a top 10 model on leaderboards. This feels like a regression from Gemini 3.5 Flash and honestly, it is one of the weakest models I have tested in the past few months. Has anyone else tested Gemini 3.6 Flash yet, and are you seeing the same thing?

🚨Gemini 3.6 Flash is trash I tested it on a 3D Golden Gate Bridge, and the results were awful. • I had to re-prompt it three times because it repeatedly ignored the instructions. • First attempt, instead of creating the requested .html file, it first tried to build the experience inside the Gemini app using simulations. • Then second attempt it started placing images from the web into the chat rather than actually producing the file. • Even after getting it to complete the task, the final output was dramatically worse than Gemini 3.1 Pro, which is 5 months old and now not even a top 10 model on leaderboards. This feels like a regression from Gemini 3.5 Flash and honestly, it is one of the weakest models I have tested in the past few months. Has anyone else tested Gemini 3.6 Flash yet, and are you seeing the same thing?

72,529 views

🚨Gemini 4 Pro (Compact Voxel Pagoda) This is from the second Gemini 4 Pro checkpoint It’s incredibly quick too, this only took around 5 minutes to generate

🚨Gemini 4 Pro (Compact Voxel Pagoda) This is from the second Gemini 4 Pro checkpoint It’s incredibly quick too, this only took around 5 minutes to generate

24,819 views

🚨GPT-6 Astra Max vs Opus 5.5 Extra (3D Seaplane) Both look absolutely incredible, Opus 5.5 wins imo I will share the scores for all the models I tested + their variants soon

🚨GPT-6 Astra Max vs Opus 5.5 Extra (3D Seaplane) Both look absolutely incredible, Opus 5.5 wins imo I will share the scores for all the models I tested + their variants soon

13,323 views

Grok 4.7 (Colosseum test) Very impressed with this one ngl I didn't test 4.7 much as we had some ridiculous drops the same week, but it's surprisingly strong with 3D stuff

Grok 4.7 (Colosseum test) Very impressed with this one ngl I didn't test 4.7 much as we had some ridiculous drops the same week, but it's surprisingly strong with 3D stuff

13,678 views

Videos

LIVE
1.2k

Anya Rossi

sweetdream.ai

SweetDream.aiSponsored•Livecam
Streaming Now

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

HD live stream
Exclusive private shows
1.2k viewers online

Free preview available • Premium content

Gemini 4 Hydrogen is actually so fast Here is a Minecraft clone done in a couple minutes
1:34
LuminaBench's profile picture

Gemini 4 Hydrogen is actually so fast Here is a Minecraft clone done in a couple minutes

Lumina

16,937 views • 1 day ago

Gemini 4 Hydrogen 3D Helicopter This is one of the fastest models I have ever seen genuinely under 2 minutes for this, it's obviously not perfect but it's only a flash tier model I will share more outputs over today
0:35
LuminaBench's profile picture

Gemini 4 Hydrogen 3D Helicopter This is one of the fastest models I have ever seen genuinely under 2 minutes for this, it's obviously not perfect but it's only a flash tier model I will share more outputs over today

Lumina

16,733 views • 1 day ago

Another Gemini 4 Hydrogen output A full V12 Engine with X-ray/wireframe, exploded and assembly views, all these new Gemini 4 models are incredibly good at 3D and are very strong at engineering and design This isn't too far off what Argon is able to do but is infinitely faster, this took no joke like a minute, it basically feels like what Ultrafast is meant to be on ChatGPT models
1:20
LuminaBench's profile picture

Another Gemini 4 Hydrogen output A full V12 Engine with X-ray/wireframe, exploded and assembly views, all these new Gemini 4 models are incredibly good at 3D and are very strong at engineering and design This isn't too far off what Argon is able to do but is infinitely faster, this took no joke like a minute, it basically feels like what Ultrafast is meant to be on ChatGPT models

Lumina

14,899 views • 1 day ago

Kimi K3.1 looks incredibly strong This is a 3D helicopter with takeoff land, 3D assembly and exploded views, X-ray and wireframe modes and a ton of detail This was made in under 15 minutes too
1:10
LuminaBench's profile picture

Kimi K3.1 looks incredibly strong This is a 3D helicopter with takeoff land, 3D assembly and exploded views, X-ray and wireframe modes and a ton of detail This was made in under 15 minutes too

Lumina

132,921 views • 16 days ago

Kimi K3.1 vs Qwen 4: 3D helicopter Same prompt, takeoff/land, 3D assembly and exploded views, very similar gen time at around 15-20 minutes for both China are not playing games anymore, these are very strong
0:56
LuminaBench's profile picture

Kimi K3.1 vs Qwen 4: 3D helicopter Same prompt, takeoff/land, 3D assembly and exploded views, very similar gen time at around 15-20 minutes for both China are not playing games anymore, these are very strong

Lumina

124,726 views • 16 days ago

🚨 Gemini 4 Pro is crazy (PS5 SVG example) This might genuinely be one of the craziest outputs I’ve got from an AI model yet For context this took around 10 minutes to complete too We are in for a treat when this thing fully drops
0:44
LuminaBench's profile picture

🚨 Gemini 4 Pro is crazy (PS5 SVG example) This might genuinely be one of the craziest outputs I’ve got from an AI model yet For context this took around 10 minutes to complete too We are in for a treat when this thing fully drops

Lumina

367,550 views • 1 month ago

🚨 Gemini 4 Pro (Xbox controller SVG) This was from "gemini 3.7 flash" in arena btw which was routing to 4 pro, it took around 20 minutes and looks crazy This example was from last night, today I can’t seem to get anything close to this
0:20
LuminaBench's profile picture

🚨 Gemini 4 Pro (Xbox controller SVG) This was from "gemini 3.7 flash" in arena btw which was routing to 4 pro, it took around 20 minutes and looks crazy This example was from last night, today I can’t seem to get anything close to this

Lumina

194,621 views • 1 month ago

🚨 Gemini 4 Pro (Colosseum) This is from the third checkpoint in Arena rn, the style has massively improved since checkpoint 1 imo, this feels a lot more polished This is a pain to get in arena so I will post more examples when I get them, I will try to get comparisons to previous checkpoints / other models too
0:24
LuminaBench's profile picture

🚨 Gemini 4 Pro (Colosseum) This is from the third checkpoint in Arena rn, the style has massively improved since checkpoint 1 imo, this feels a lot more polished This is a pain to get in arena so I will post more examples when I get them, I will try to get comparisons to previous checkpoints / other models too

Lumina

98,381 views • 1 month ago

Oh my GOD Kimi K3.1 is Broken One shot fully working 3D Train with speed slider, assembly / exploded views, wireframe and stat tracking Wtf is going on with these Chinese models, this feels design wise similar to a GPT model but just has that something extra like Anthropic models on top
0:59
LuminaBench's profile picture

Oh my GOD Kimi K3.1 is Broken One shot fully working 3D Train with speed slider, assembly / exploded views, wireframe and stat tracking Wtf is going on with these Chinese models, this feels design wise similar to a GPT model but just has that something extra like Anthropic models on top

Lumina

42,415 views • 16 days ago

Opus 5.5 vs Sonnet 5.5 vs Astra vs 6.1 Sol (Mario Kart test) All on the same prompt Sonnet used over 2x as many tokens as Opus in the same amount of time and imo looks the best of all four
0:54
LuminaBench's profile picture

Opus 5.5 vs Sonnet 5.5 vs Astra vs 6.1 Sol (Mario Kart test) All on the same prompt Sonnet used over 2x as many tokens as Opus in the same amount of time and imo looks the best of all four

Lumina

61,680 views • 26 days ago

Qwen 4 (Kalmia) 3D Drivable Car I'm honestly not sure which model this is from the Qwen lineup, and it still has some issues But this is very cool and only took around 20 minutes to generate
0:41
LuminaBench's profile picture

Qwen 4 (Kalmia) 3D Drivable Car I'm honestly not sure which model this is from the Qwen lineup, and it still has some issues But this is very cool and only took around 20 minutes to generate

Lumina

37,011 views • 16 days ago

This is Kimi K3.1's voxel pagoda vs frontier models I also did GPT-6 Astra so you can see what I mean about it having similar design style, I attached carbon too from checkpoint 1 i'm redoing it now with the new checkpoint The actual tower is near identical in design to Astra, and you know what that means 🙏
0:27
LuminaBench's profile picture

This is Kimi K3.1's voxel pagoda vs frontier models I also did GPT-6 Astra so you can see what I mean about it having similar design style, I attached carbon too from checkpoint 1 i'm redoing it now with the new checkpoint The actual tower is near identical in design to Astra, and you know what that means 🙏

Lumina

34,175 views • 16 days ago

Gemini 4 Carbon Voxel Pagoda Carbon has been in arena since yesterday, so I will show off some outputs I got over today It's a noticeable improvement over Argon but is incredibly slow, like 2-4x slower depending on task
0:27
LuminaBench's profile picture

Gemini 4 Carbon Voxel Pagoda Carbon has been in arena since yesterday, so I will show off some outputs I got over today It's a noticeable improvement over Argon but is incredibly slow, like 2-4x slower depending on task

Lumina

35,371 views • 17 days ago

Gemini 4 Carbon vs Opus 5.5: 3D Helicopter Carbon was done in arena in a tad over 30 minutes, Opus 5.5 is on Extra and took 55 minutes Something to keep in mind is that Opus 5.5 is in its preferred harness, Claude Code, I personally prefer Carbon's design but Opus 5.5 is more polished overall
1:37
LuminaBench's profile picture

Gemini 4 Carbon vs Opus 5.5: 3D Helicopter Carbon was done in arena in a tad over 30 minutes, Opus 5.5 is on Extra and took 55 minutes Something to keep in mind is that Opus 5.5 is in its preferred harness, Claude Code, I personally prefer Carbon's design but Opus 5.5 is more polished overall

Lumina

28,896 views • 14 days ago

Gemini 4 Carbon (Voxel Pagoda) vs Frontier Models These are some frontier models against it for reference, it has noticeably higher quality than Argon does, as i said it's a lot slower though
0:27
LuminaBench's profile picture

Gemini 4 Carbon (Voxel Pagoda) vs Frontier Models These are some frontier models against it for reference, it has noticeably higher quality than Argon does, as i said it's a lot slower though

Lumina

33,514 views • 17 days ago

Gemini 4 Carbon: 3D Helicopter Kimi K3.1 Pro / Qwen 4 were good, but this just has a completely different level of 3D and engineering understanding I will do a comparison with Opus 5.5 soon, but this was one shot in arena in a little over 30 minutes
1:37
LuminaBench's profile picture

Gemini 4 Carbon: 3D Helicopter Kimi K3.1 Pro / Qwen 4 were good, but this just has a completely different level of 3D and engineering understanding I will do a comparison with Opus 5.5 soon, but this was one shot in arena in a little over 30 minutes

Lumina

25,296 views • 14 days ago

Qwen 4 Jet engine This is from Kalmia checkpoint 2, very strong with 3D, physics and engineering again From my testing the other day, Qwen 4 Max is better than Kimi K3.1 Pro too
0:54
LuminaBench's profile picture

Qwen 4 Jet engine This is from Kalmia checkpoint 2, very strong with 3D, physics and engineering again From my testing the other day, Qwen 4 Max is better than Kimi K3.1 Pro too

Lumina

19,421 views • 14 days ago

Gemini 4 Carbon is incredible at making games Checkpoint 1 is testing in arena currently and this is one of the cleanest outputs I have got from the Super Mario prompt Gen time was quite high like 30 minutes for this but it's no longer lazy like Argon, it just gets what you want and does it perfectly first try
1:14
LuminaBench's profile picture

Gemini 4 Carbon is incredible at making games Checkpoint 1 is testing in arena currently and this is one of the cleanest outputs I have got from the Super Mario prompt Gen time was quite high like 30 minutes for this but it's no longer lazy like Argon, it just gets what you want and does it perfectly first try

Lumina

22,230 views • 17 days ago

Gemini 4 Argon (3D Seaplane on Water) This might be the best out of all the models i tested this prompt on, beautiful
1:07
LuminaBench's profile picture

Gemini 4 Argon (3D Seaplane on Water) This might be the best out of all the models i tested this prompt on, beautiful

Lumina

23,900 views • 24 days ago

🚨 Gemini 4 Argon vs Opus 5.5 (Voxel Pagoda) Gemini 4 Argon is from arena during the weeks it was in testing, Opus 5.5 is on high here too For more context Gemini 4 was around twice as fast too
0:21
LuminaBench's profile picture

🚨 Gemini 4 Argon vs Opus 5.5 (Voxel Pagoda) Gemini 4 Argon is from arena during the weeks it was in testing, Opus 5.5 is on high here too For more context Gemini 4 was around twice as fast too

Lumina

30,754 views • 1 month ago