Nipuna

🚨 European AI labs are soo back 🔥 Mistral just released the open weights for Le Chonk (Mistral Large 4). Here's everything you need to know: • 1-trillion-parameter, natively multimodal MoE: Combines fast chat and deep reasoning in a single model. • #1 open-weight coding model: Ranked first on Cognition’s FrontierCode benchmark. • Top token efficiency: Artificial Analysis reports that it delivers frontier-level intelligence using far fewer tokens. • State-of-the-art open model for cybersecurity: Already validated by CrowdStrike, Cloudflare, and Zscaler. • Fully open weights: Download it, fine-tune it with Mistral Forge, and self-host it anywhere. • Built for sovereign AI: Trained entirely on European infrastructure. The weights are live on Hugging Face.

The new 𝕏 Money specs are actually wild: 🤯 • Up to 3% cashback on the new silver 𝕏 Card • Up to 6.00% APY with no minimum balance • Up to $10M in FDIC insurance Crazy times we live in.

🚨 DeepSeek is preparing to drop its V5 lineup. New code spotted in deepseek-harness lists 3 upcoming models: • deepseek-v5 • deepseek-v5-open • deepseek-v5-next Looks like we're getting more DeepSeek models before the end of the year.

OpenAI is merging ChatGPT "Chat" and "Work" into a single interface. OpenAI's Tibo just confirmed it's coming later this year.

490 followers in just 4 days... 🤯 𝕏 has the best community of people. Thank you everyone 🙌 If you ❤️ AI, let's connect!

Apple just introduced the new MacBook Pro. It's the first MacBook with a touchscreen, and they brought the glowing Apple logo back after 10 years. Apple is cooking fr. 🔥

🚨 Google just introduced Gemini 4 Carbon. The biggest thing on this model is state-of-the-art cybersecurity and long-horizon agentic coding. Here's everything you need to know: Gemini 4 Carbon has: • #1 rank on the Artificial Analysis Intelligence Index v5.0 (score of 59) • Massive cybersecurity gains: 88.2% on SEC-bench Pro, 54.8% on ExploitGym, and 71.0% on CWE-bench • 2x+ better secure code generation than Gemini 4 Argon (and 10x+ more high-quality patches) • #4 on WebDev Arena (+89 Elo to hit 1,769) and built for multi-day agentic engineering • 0.0% indirect prompt injection success, 0.0% cheating on Terminal-Bench 4.0, 2x fewer unsupported claims, and 80% less high-severity misalignment Introductory API pricing starts at $2/1M input and $10/1M output tokens (95% off cached inputs), before moving to $4/$20 later. Rolling out this week to cyber defenders via the Fairwind Program, with broader developer and consumer access following U.S. government pre-release review.

How AI labs name their models: • Anthropic (Literature): Haiku, Sonnet, Opus, Mythos • OpenAI (Astronomy): Luna, Terra, Sol, Astra • Google (Chemistry): Argon, Carbon, Hydrogen • Kimi (Music): Adagio, Andante, Allegro • DeepSeek (Calendar): 0324, 0528, 1017 • SpaceXSI (Sci-Fi): Grok, Tensor, Quasarflux

𝕏 just launched Photo Resharing. Here's how it works: • Long-press any photo you want to share • Tap "Post photo" • Automatic credit badge goes on the image • Both you AND the original creator earn from Original Content Rewards Massive W for original creators.

Another 153 followers in 24 hours... 🤯 𝕏 has the best community on the internet. If you ❤️ AI, let's connect!

OpenAI introduced GPT-6.1 Terra. Here's everything you need to know: • The biggest thing on this model is efficiency. • 50% cheaper API than GPT-5.6 Terra & GPT-6.1 Sol • Matches GPT-6 Astra at ~1/10th the price per token • 3× faster task completion than Astra & Sol • Stronger computer use (reads screens and acts across apps) • Big jump on Agents’ Last Exam for real work tasks • Factual errors dropped from 53.6% to 16.7% at low effort Use Terra for everyday work and agent workflows. Stick to Sol or Astra for deep research and heavy coding.

Google just dropped Genie 4.🔥 Here's everything you need to know: • Real-time interactive world model with unprecedented realism. • Native audio generation, including ambient sounds, music, and interactive dialogue, all generated alongside the world. • Real-time natural language control, allowing users to change the weather, move the camera, and add characters or objects. • Improved physics and more natural human movement. • Built on Gemini Omni and other Gemini models to improve coherence, continuity, and multi-step reasoning. • Designed for safe simulation in robotics and AI agents, as well as creative applications such as filmmaking, gaming, and prototyping. Genie 4 is currently available as a research preview for trusted testers. Broader access is coming to Google AI Ultra subscribers through Project Genie, along with early access for partners via the Gemini API and Google AI Studio. All generated content includes SynthID watermarking.

Normal people buying an iPhone Duo: 🕺💃 Me the second I unbox an iPhone Duo:

Bro disappeared like he never existed 😭

Hey @grok pick the most useless model from the GPT-6 family and nuke it off this grid. Don't hold back.

183 followers in 1 night... 🤯 𝕏 has the best community of people. If you ❤️ AI, let's connect!

Genuine question: Do you guys swear at your Grok bot, Meta Muse, or ChatGPT Dots when you get mad?

What comes after AI? 🤔

Be honest. Are you taking the $100,000 cash or unlimited tokens for life?

🚨🚨🚨 Oh wow, ChatGPT just changed the usage limits to a monthly cycle. That's craaaazzzyyyy!!!