Just completed generating GGUF quantized versions of R1 1776 - an uncensored variant of DeepSeek R1 from Perplexity Aravind Srinivas Original model size(16bit): ~1.35TB Quantized variants: from 244GB (Q2) to 714GB (Q8) Have fun! huggingface.co/tensorblock/r1…

Successfully tested running full-size Deepseek 70B model on three 8x NVIDIA RTX 3080 rigs, achieving 25 tokens/s through 3-way pipeline and 8-way tensor parallelism optimization. Each rig is equipped with 8x 10GB consumer GPUs (typical crypto mining rig configuration),

QwQ-32B (Qwen ) model sharding between M1 MacBook Pro (16GB) + RTX 4060 Ti. Enabling efficient inference through model quantization and cross-device parallel computing. Demonstrating production-ready performance on consumer hardware. Technical benchmarks coming soon.

Hello April. 🚧 Something exciting is brewing. We’re unlocking the next chapter. Not just one drop — a wave. Stay tuned. It’s almost time to mine thyself.

We’re tackling a key challenge in decentralized AI: efficient inference verification. Current methods = costly & redundant. Our solution: Proof of Cache – lightweight, deterministic, and scalable. 🔍 Check out how TensorBlock is making decentralized AI practical.

We're proud to announce that TensorBlock is now officially part of the #NVIDIA Inception Program! This partnership strengthens our mission to build the future of democratic and composable intelligence. #NVIDIAInception #AI #Startup #LLM #TensorBlock

🚀 Excited to be a launch partner for Nautilus, powering verifiable AI inference with tamper-proof TEEs. Through our Proof of Cache protocol, every AI response on TensorBlock carries cryptographic provenance — trust built into the compute. More to come. 🔗 TensorBlock |

Thank you AMD for an incredible event and all the exciting efforts behind it. It was a pleasure to share our insights on open-source AI and the emerging era of AI agents during the workshop time. We’ve always believed in AMD’s strengths, especially in the inference space

🚨 TensorBlock Forge is now live on Product Hunt! One API, all your AI models. OpenAI-compatible, privacy-first, and prod-ready. Help us hit the front page by leaving a comment and upvote! producthunt.com/products/tenso…

finally got some real sleep after 24h straight in the office launching Forge on Product Hunt 😸 😵‍💫 ft. my loyal teammate Pikachu & our office “bed” 🫠 hotfixing till 5am, answering Discord, support tickets, user calls, adhoc meetings… last time I felt this was Libra launch at

Big milestone! We just hit 3 million downloads on Hugging Face 🎉 With 4,823+ quantized models, each supporting 10+ precision formats, we're committed to making open-source AI more accessible and efficient for everyone. And we’re far from done — we’ll keep building and