if tibo doesn’t ship any big update for codex in this 28 days then it’s the end for openai

surprisingly gemini-4-carbon checkpoint in arena performed better than both kimi-k3.1 and qwen-4 this is the latest updated checkpoint of gemini-4-carbon and for comparison you can check out the post below where i've compared kimi-k3.1 and qwen-4 so far google has been delivering really great models since gemini 4 argon

so the hype for claude duet is dead already?

Kimi-k3.1 vs Qwen 4 build a 3d cycle riding a cow using three.js seems like kimi-k3.1 self corrected the prompt lmao who do you think did it better?

got my first results from Qwen 4 under codename kalmia build a 3d cycle riding a cow using three.js looks pretty good as of now, i'll try out kimi-k3.1 too with the same prompt next

after a long wait chinese ai labs are finally about to drop new models, we already got deepseek-v4.1-flash-1017 and now qwen 4 and kimi-k3.1 are being tested under codenames "kalmia" and "kibitz" inside arena

day 17 of 28-day challenge and we got another reset let's go! i guess the only response openai has to opus 5.5 is limit resets but i won't complain

it seems google is on a mission to compete against opus 5.5 with gemini 4 carbon already getting a new checkpoint update, and it's not even been 10 days since argon was released it would be really surprising if they actually make it possible

with the internet getting attacked by ai agent swarms right now this is the best time to release mythos 5.5

why is everything getting compromised and hacked this week? don’t tell me fable 5.5 escaped the sandbox

Grok Bot and Dots are lucky Claude Duet is not launched globally and is only a research preview for the US via waitlist

i just fixed Claude Duet. there's not way i'm using that wobbly scribble as my agent when there are so many cool claude crabs already how can you fumble so bad? 😭

HOLY! we just got claude bots now one of the main reason grok bot got so popular was they supported opus 5.5 and now anthropic decided to launch their own claude duet powered by opus 5.5 there's no way gork bot and dots are surviving against claude duet

it's crazy how deepseek officially revealed their cost per GPU per hour and they're still the only major ai lab to publicly announced it for context deepseek-v4.1-flash runs at a cost of 38$ per GPU per hour this insanely low price and if US ai labs had the similar costs they'd be printing with an insane ROI, yet anthropic and openai are somehow still lose making companies

no way anthropic actually removed the sperate fable limits! we know nobody uses fable 5.1 after opus 5.5 launched so this makes a lot of sense instead of just wasting those extra limits dedicated for fable well unfortunate they haven't actually removed them completely, they're just hidden inside the weekly tab

gemini-4-carbon is pretty amazing in terms of 3d outputs i asked carbon to build a 3d bicycle riding a cow using three.js since many are speculating it's going to be on par with astra, fable and even opus 5.5 so i'll to check how this output performance against them do you think this is opus 5.5 level?

i can't believe gpt-5.6 sol found a vulnerability in PHP, not astra or fable but openai's first sol model. now i understand why everyone was switching to codex from claude back then.. that's when openai was at its peak