DeepSeek-V3.1 from DeepSeek is now live on AkashChat! Run Open Source models on the Open Cloud, free to access, free to explore... The way AI should be. Try it here: chat.akash.network/models/DeepSee…

.xAI’s Grok 2 weights have been released on Hugging Face huggingface.co/xai-org/grok-2

👀

Apparently if you ask ChatGPT 5 how many strawberries are in the letter R, it goes right back to its old self.

🔥

🔥 New on AkashChat: Qwen3-Next-80B-A3B is live. An advanced Mixture-of-Experts model with 80B total parameters and just 3B activated per token—delivering massive efficiency gains. Featuring Hybrid Attention (Gated DeltaNet + Gated Attention), it unlocks 10× faster inference on

Thrilled to report Decentralized Training has arrived on Akash. Training AI on globally distributed heterogeneous compute allows you to use GPUs from anywhere in the world to train AI, which is key to scaling AI and solving the energy crisis created by Big Cloud. We at Akash

It's already starting: Google is officially purchasing nuclear power. Bloomberg is now reporting that NextEra Energy, $NEE, plans to restart a nuclear power plant in Iowa, primarily to supply Google data centers. The 615-megawatt plant shut down in 2020, well before the AI

Seeing akashml.com perform flawlessly while cruising at 30,000 feet makes me a happy man. Akash Chat API users will get access to the platform starting Monday.

AkashML is live! A fully managed AI inference service on Akash Network. Deploying open models like DeepSeek shouldn’t cost $50K/month or require complex setup. AkashML makes inference simple, affordable, and automatic, so developers can run AI workloads on decentralized GPUs

Massive thanks to forces behind Akash: → Our community for supporting and building with us from day one → Our providers for making GPU compute accessible for all → The devs for shipping every day to make decentralized compute the new standard for the global economy Happy

🚀 Launching DeepSeek-V3.2 & DeepSeek-V3.2-Speciale — Reasoning-first models built for agents! 🔹 DeepSeek-V3.2: Official successor to V3.2-Exp. Now live on App, Web & API. 🔹 DeepSeek-V3.2-Speciale: Pushing the boundaries of reasoning capabilities. API-only for now. 📄 Tech

Yann LeCun says the best open models are not coming from the West. Researchers across the field are using Chinese models. Openness drove AI progress. Close access, and the West risks slowing itself.

MiniMax-M2.5 is now open source. Trained with reinforcement learning across hundreds of thousands of complex real-world environments, it delivers SOTA performance in coding, agentic tool use, search, and office workflows. Hugging Face: huggingface.co/MiniMaxAI/Mini… GitHub:

🚀 Qwen3.5-397B-A17B is here: The first open-weight model in the Qwen3.5 series. 🖼️Native multimodal. Trained for real-world agents. ✨Powered by hybrid linear attention + sparse MoE and large-scale RL environment scaling. ⚡8.6x–19.0x decoding throughput vs Qwen3-Max 🌍201