Legit

Made by Veo 3 "A squirrel attempts to launch a homemade rocket into space"

Guess the model on VoxelBench!

So Mythos was, indeed, not marketing hype. Remember this is a general purpose model that just happens to be good at finding exploits because good models are good at lots of things. Expect similar from OpenAI & Google. And from open models in 8 months. hacks.mozilla.org/2026/05/behind…

GPT-5.5 Pro has ranked 1st on VoxelBench It scores 100+ Elo points higher than GPT-5.5!

GPT-5.5 Pro is now on VoxelBench! as always, there are hundreds of generations Next we'll run these models on our multimodal set, see you soon!

GPT-5.5 has ranked 1st on VoxelBench It's the first model to cross a score of 2000! Its Win Rate is currently 96% based on 517 votes so far from our users

GPT-5.5 xHigh has landed on VoxelBench! this model has made some amazing voxel art "a skyline of New York city as viewed from the Hudson river"

IT'S HAPPENING!

Exciting news - GPT-Image-2 by @OpenAI has claimed the #1 spot across all Image Arena leaderboards! A clean sweep with a record-breaking +242 point lead in Text-to-Image - the largest gap we’ve seen to date. - #1 Text-to-Image (1512), +242 over #2 (Nano-banana-2 with web-search aka gemini-3.1-flash-image) - #1 Single-Image Edit (1513), +125 over #2 (Nano-banana-pro aka gemini-3-pro-image) - #1 Multi-Image Edit (1464), +90 over #2 (Nano-banana-2) No model has dominated Image Arena with margins this wide. Huge congratulations to @OpenAI on this major breakthrough in image generation! More performance breakdowns by category in the thread below.

@chetaslua @TheRealAdamG wow

Holy SHIIIIIIITTTTTTTTT @TheRealAdamG what did you guys created , this is insane GPT pro update created a masterpiece under 11 mins, wtf holy shit prompt and codepen link in comment

New info seemingly from DeepSeek staff - A new, much larger base model will be released soon - DeepSeek V3.2 seems to be a larger model than the one currently deployed on the web I used Gemini for translation

@OfficialLoganK need multi user vibe coding in build mode why - so entire teams can iterate on the same web app

Our AI Studio vibe coding roadmap for the new few weeks: - Design mode - Figma integration - Google Workspace integration - Better GitHub support - Planning mode - Immersive UI - Agents - Multiple chats per app - Simplified deploys - G1 support And more, should be fun : )

Yes, Gemma 4 will soon be released this time their largest size might be around 120B total with 15B active

@gdb need more posts like this Greg, ty

Benchmarks? Where we’re going, we don’t need benchmarks.

@Meadowbrook_ being "pretty good" does not negate the fact that it wasn't Omnimodal they released AVM months later and native image gen came a year later even then, to use any of these capabilities it would route you to a different model even today in the API they are different models lol

@legit_api Nah. It was pretty good for what it was in 2024, most models until recently had 128K as the standard, Gemini was the outlier, even Opus 4.5 had 200K iirc.

OpenAI is developing a new Omni model Omni models are fully multimodal Input: Text + Image + Video + Audio Output: presumably any mix of above It's what the original GPT-4o was supposed to be, but never was

excuses, OpenAI is a frontier lab, the one that massively accelerated AI development with ChatGPT Google shipped 1M tokens context window with Gemini 1.5 Pro in Feb 2024 (2 years ago!) The release of 4o was very half baked, remember all the delays for AVM + native image gen The released 4o was never Omnimodal

@legit_api It was to the best of its abilities, but the modules weren't there and there's only so much you can do with 128K tokens.

GPT-5.4 has landed on the Arena unclear which variant

Nano Banana 2 vs Nano Banana Pro "A pixelated game-style background, a dark color scheme, and minimal patterns"