🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

We'd be happy to see it hosted by an Inference Provider via huggingface. Upvoted

Brain surgery like this is what keeps Yud up at night

DeepSeek-R1T-Chimera，结合 DeepSeek R1 的智能性和 V3 的token 效率，由 TNG Technology Consulting GmbH 团队开发 主要特点 - 规模：拥有 685B 参数，属于超大规模模型 - 类型：Text Generation Transformers - 架构：基于 DeepSeek-MoE Transformer 架构 技术特点 - 模型合并项目，将 DeepSeek-R1 和 DeepSeek-V3(0324)

Oh man, lucky day 😅 R1T-Chimera is ranked the #2 trending model on OpenRouter Sure, "trending" is a temporary attention metric, not reflecting total usage. And the world spins fast: everybody talks about Qwen3 now. Still a nice screenshot with Google AI, Microsoft,

THE WHALE IS BACK!!! 🐳🐳🐳 huggingface.co/deepseek-ai/De…

DeepSeek uploaded a new model on huggingface: DeepSeek-Prover-V2 It seems, the architecture is identical to V3 and R1 models, because: model_config.py shows no difference, also the safetensor index files are the same. One minor diff is a new experimental feature in

There's a new paper circulating looking in detail at LMArena leaderboard: "The Leaderboard Illusion" arxiv.org/abs/2504.20879 I first became a bit suspicious when at one point a while back, a Gemini model scored #1 way above the second best, but when I tried to switch for a few

BOOOOM! you can now use the latest DeepSeek Prover V2 directly on the model page powered by Novita AI 🔥 Open Source FTW! 💥

x.com/tngtech/status…

Eight new AMD MI325X GPUs joined our compute cluster of NVIDIA H100s. The new Supermicro server is an AI machine with spectacular 2 Terabytes of total GPU memory in one ~10kW node. ROCm worked right away with full VRAM and GPU utilization, allowing new types of

Hello #USA 🇺🇸 TNG Technology Consulting USA Inc. is now incorporated in #Austin, #Texas. Thanks to our existing clients in #SiliconValley and #NewYork. We look forward to meeting more interesting people, fast companies and hard #IT problems to solve.

x.com/i/article/1928…

More evidence for the effectiveness of the Chimera construction method: Taking DeepSeek's R1-0528 release, we started benchmarking new Chimera variants on AIME-24 and SimpleQA. R1-0528 significantly improves AIME performance from 79.8 to 91.4 while doubling the amount of output