Lentils's banner

Lentils

@Lentils80 • 11,922 subscribers

Early AI news, leaks and cool tests. Reach me via DMs or [email protected] Discord server: https://t.co/nektADFNBS

Shorts

Google is preparing to release an updated version of Omni Flash (models/gemini-omni-1.1-flash) The underlying model id is "omni-1p2", indicating that Google is using the Omni Flash 1.2 checkpoint for 1.1 The attached video is an output from it, generated in 43 seconds. I wish they wouldn't release it at all atp...

Google is preparing to release an updated version of Omni Flash (models/gemini-omni-1.1-flash) The underlying model id is "omni-1p2", indicating that Google is using the Omni Flash 1.2 checkpoint for 1.1 The attached video is an output from it, generated in 43 seconds. I wish they wouldn't release it at all atp...

44,796 views

Videos

Anya Rossi

sweetdream.ai

SweetDream.ai•Sponsored•Livecam

Watch Anya Live

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

Exclusive private shows

1.2k viewers online

Private Show

Join now for exclusive access

Free preview available • Premium content

🚨 Major Scoop: OpenAI has just expanded internal testing of GPT Astra, codenamed "mozaik-alpha-fdm", signifying release might be near AND of course like previous times, we have the FIRST EVER public outputs of it for y'all😉 Both outputs are zero-shot on Max effort. Frontend *might* have finally been fixed, and the model gives a lot of attention to details

🚨 Major Scoop: OpenAI has just expanded internal testing of GPT Astra, codenamed "mozaik-alpha-fdm", signifying release might be near AND of course like previous times, we have the FIRST EVER public outputs of it for y'all😉 Both outputs are zero-shot on Max effort. Frontend might have finally been fixed, and the model gives a lot of attention to details

764,948 views • 1 month ago

Gemini 4 Pro checkpoints have finally started appearing internally a few days ago. This is the first ever output from the model, internally codenamed "argon". It took 2.4 minutes on High thinking effort. It has a 256k token "output limit", compared to 64k in previous Gemini models. I also heard it will ship with a 2M context window, tho it's still not decided if they'll do it.

Gemini 4 Pro checkpoints have finally started appearing internally a few days ago. This is the first ever output from the model, internally codenamed "argon". It took 2.4 minutes on High thinking effort. It has a 256k token "output limit", compared to 64k in previous Gemini models. I also heard it will ship with a 2M context window, tho it's still not decided if they'll do it.

393,905 views • 26 days ago

Seeing as Claude Mythos is releasing soon, I have two VERY astonishing outputs to share from it. 👀 ZERO-SHOT and LOW effort as well! These are the best outputs I've seen for this prompt ever since the October 2025 Gemini A/B models.

Seeing as Claude Mythos is releasing soon, I have two VERY astonishing outputs to share from it. 👀 ZERO-SHOT and LOW effort as well! These are the best outputs I've seen for this prompt ever since the October 2025 Gemini A/B models.

587,072 views • 4 months ago

Claude Opus 5 made this beautiful painterly world in a single HTML file, everything made procedurally It implemented a realistic grass and wind field simulation so that the millions of grass blades on-screen react to the wind

Claude Opus 5 made this beautiful painterly world in a single HTML file, everything made procedurally It implemented a realistic grass and wind field simulation so that the millions of grass blades on-screen react to the wind

331,188 views • 2 months ago

More pre-release GPT Astra outputs for my distinguished audience Both outputs were made in a custom C++20 engine, utilizing a custom Vulkan renderer, without using an off-the-shelf engine or imported assets The Kaiju game has fully simulated crowds with reactive smoke and fire (not static) , and the zombie game can render thousands of zombies on-screen

More pre-release GPT Astra outputs for my distinguished audience Both outputs were made in a custom C++20 engine, utilizing a custom Vulkan renderer, without using an off-the-shelf engine or imported assets The Kaiju game has fully simulated crowds with reactive smoke and fire (not static) , and the zombie game can render thousands of zombies on-screen

162,777 views • 1 month ago

Terraria recreation by pre-release GPT Astra, made in a single HTML file in one user turn with no imported assets It included multiple bosses as well as a transition to hard mode

Terraria recreation by pre-release GPT Astra, made in a single HTML file in one user turn with no imported assets It included multiple bosses as well as a transition to hard mode

149,756 views • 1 month ago

Another impressive and early Claude Mythos output for y'all. 😀 I literally just told it to generate a macOS clone and to do its absolute best. It generated 50k tokens (3k lines of code), made a fully functional browser, made a music player with generated songs and added a lot of sneaky witty details. AND this is also LOW effort!

Another impressive and early Claude Mythos output for y'all. 😀 I literally just told it to generate a macOS clone and to do its absolute best. It generated 50k tokens (3k lines of code), made a fully functional browser, made a music player with generated songs and added a lot of sneaky witty details. AND this is also LOW effort!

427,780 views • 4 months ago

Some of the first Gemini 3.6 Flash outputs for y'all, and yeah... Genuinely the worst results I've ever gotten on these two prompts. Terrible frontend capability (2 shot btw) and spatial reasoning, at least it's very fast ig I cope this isn't actually running on High thinking level but idk

Some of the first Gemini 3.6 Flash outputs for y'all, and yeah... Genuinely the worst results I've ever gotten on these two prompts. Terrible frontend capability (2 shot btw) and spatial reasoning, at least it's very fast ig I cope this isn't actually running on High thinking level but idk

198,272 views • 2 months ago

I compared Claude Fable 5 to GPT-5.5 in this Power Rangers prompt Thing is, Fable 5 is using Low thinking effort and GPT-5.5 is using xhigh Safe to say, the results are... not even close. 5.5's output is bad across the board, from the UI to the actual voxel scene itself🥲 1st video: Claude Fable 5 (Low effort) 2nd video: GPT-5.5 (xhigh)

I compared Claude Fable 5 to GPT-5.5 in this Power Rangers prompt Thing is, Fable 5 is using Low thinking effort and GPT-5.5 is using xhigh Safe to say, the results are... not even close. 5.5's output is bad across the board, from the UI to the actual voxel scene itself🥲 1st video: Claude Fable 5 (Low effort) 2nd video: GPT-5.5 (xhigh)

207,512 views • 4 months ago

More outputs from "claude-marshmallow-eap" and "claude-melon-eap" Both got updated yesterday. After further testing, "melon" actually looks to be a Fable checkpoint while "marshmallow" feels like a weaker Fable or strong Opus checkpoint Release is expected to be as soon as this week, as both models got removed a few hours ago

More outputs from "claude-marshmallow-eap" and "claude-melon-eap" Both got updated yesterday. After further testing, "melon" actually looks to be a Fable checkpoint while "marshmallow" feels like a weaker Fable or strong Opus checkpoint Release is expected to be as soon as this week, as both models got removed a few hours ago

64,933 views • 1 month ago

One of the first Kimi K3 outputs for y'all, tried it on frontend 👀 First impressions is that it's VERY slow, even slower than Fable. This took 35 minutes to finish. However, this is one of the best outputs I've ever seen from this prompt, better than many frontier models.

One of the first Kimi K3 outputs for y'all, tried it on frontend 👀 First impressions is that it's VERY slow, even slower than Fable. This took 35 minutes to finish. However, this is one of the best outputs I've ever seen from this prompt, better than many frontier models.

104,085 views • 2 months ago

Told Claude Fable 5 to make a Vampire Survivors clone in pure C++, no engine, no asset files Everything from sprites, sounds, graphics to music tracks is generated by code at runtime Source code is in the thread below

Told Claude Fable 5 to make a Vampire Survivors clone in pure C++, no engine, no asset files Everything from sprites, sounds, graphics to music tracks is generated by code at runtime Source code is in the thread below

99,199 views • 3 months ago

🚨 Looks like a new Qwen model (3.8/4?) appeared on LMArena under the stealth name "Kaleb" It insists on introducing itself as Claude, but users identified it as Qwen by looking for specific output signals only Qwen produces It's not confirmed as Qwen but it's def a Chinese model (by asking political questions)

🚨 Looks like a new Qwen model (3.8/4?) appeared on LMArena under the stealth name "Kaleb" It insists on introducing itself as Claude, but users identified it as Qwen by looking for specific output signals only Qwen produces It's not confirmed as Qwen but it's def a Chinese model (by asking political questions)

80,375 views • 2 months ago

Gemini 3.2 got renamed to 3.5 👀 In other news, below are outputs from a strong Gemini 3.5 Pro checkpoint called "Cappuccino"

Gemini 3.2 got renamed to 3.5 👀 In other news, below are outputs from a strong Gemini 3.5 Pro checkpoint called "Cappuccino"

97,820 views • 4 months ago

Got an output from the strongest pre-release Gemini 3.5 Pro checkpoint 👀 I compared it to Claude Fable 5 and GPT-5.5 on the same exact prompt Yeah... This model is mostly dead on arrival (tried it on other prompts too ofc). It will likely be the "poor man's model" next to 3.5 flash. 1st video: Gemini 3.5 Pro (pre-release) 2nd video: GPT-5.5 3rd video: Claude Fable 5

Got an output from the strongest pre-release Gemini 3.5 Pro checkpoint 👀 I compared it to Claude Fable 5 and GPT-5.5 on the same exact prompt Yeah... This model is mostly dead on arrival (tried it on other prompts too ofc). It will likely be the "poor man's model" next to 3.5 flash. 1st video: Gemini 3.5 Pro (pre-release) 2nd video: GPT-5.5 3rd video: Claude Fable 5

50,395 views • 4 months ago

GPT-6 Astra made this blackhole renderer using C++, Vulkan and raytracing It implemented advanced Kerr spacetime math, tracing light bent from 211000 stars, frame dragging and relativistic color shifts It made it run at 1080p/~30 FPS in realtime high precision on a GTX 1050 ti

GPT-6 Astra made this blackhole renderer using C++, Vulkan and raytracing It implemented advanced Kerr spacetime math, tracing light bent from 211000 stars, frame dragging and relativistic color shifts It made it run at 1080p/~30 FPS in realtime high precision on a GTX 1050 ti

16,877 views • 1 month ago

No more content to load