🚨 Massive leak: DeepSeek-V4 benchmark evaluation is out, and the numbers are staggering. The chart claims SOTA across general, math, code, and agent tasks, outperforming expected next-gen models like GPT-5.3, Claude Opus 4.6, and Gemini 3.1 Pro Preview. Key highlights: 📐 AIME

DeepSeek is reportedly raising funds at a $10B+ valuation. Do you guys think this valuation is reasonable? 🤔

are you sure?🤔

GPT-5.5/Spud?😍

50 minute half marathon. yeah we are cooked😱

kimi k2.6 just dropped. the new background agent can run autonomously for 5 solid days. 120 hours of continuous work with zero sleep. the ultimate cyber worker is here. 💅 #kimi #aiagent #tech

funnyh😄

GPT-5.5 has come

deepseek v4 looks insane. kinda bummed it's text only and not multimodal, but still super hyped for it ngl 😭#deepseekv4 #llm #ai

need this on my codex rn. when is the rollout?

So powerful!

admitting to the lobotomy right after gpt-5.5 drops is crazy timing 💀 be honest, did u guys use 5.5 to help u find the bugs?

hope v5 brings multimodal tbh

Thank you

Gpt-5.6?🤔

looks like deepseek is quietly beta testing an image recognition mode right now. does this officially count as multimodal? 👀

the capability gap is literally widening on this chart 💀 the main W for open source rn is honestly just cost efficiency. thank god for open source tho, without them forcing the prices down, these closed-source monopolies would be charging us way more for api calls 😭

bro why is the deepseek api suddenly throwing 400 invalid user_id errors outta nowhere