Mistral AI reposted this
Let the Mistral AI MoE tokens fly 📈 🚀 #Mixtral 8x7B with NVIDIA #TensorRT #LLM on #H100. ➡️ https://nvda.ws/3L5WIZ0 ✨
Mixture of Experts (MoE) optimize query expense, better response time, reduced compute power necessary, and rewarding domain specific expertise. Use-Case: LLM trained in Cyber Security hosting an AI Security Agent, as an identity on a network. Securing IT (Information Technology), IoT (Internet of Things, drones), OT (Operational Technology), ICS (Industrial Control Systems), Critical Infrastructure, and human endpoints is too large an attack surface for human capabilities alone. AI presents new unique challenges to cyber security and solving the problem requires an AI solution. Secure AI computing is an immediate concern.
Exciting🚀Thanks for sharing💯🤠
Excellent ! However the both did not answer the request of my students 🤔 for a talk.
Amazing! Love it !!
Can't wait!
Very cool !
Very cool image
Awesome product
Innovation Specialist - Engineering | Design | AI | UX | IP | Digital & Physical Prototyping - Breaking problems into pieces.
3moMixtral 8x7B - 2x H100's - ~$5/hr - 60 tokens/s - 216 thousand tokens/hour - $23.15/million tokens ///// Groq - 575 tokens/s - $0.24/million tokens - $0.50/hr - 2.07 million tokens/hour Weird flex, but I guess what other hardware could you buy... For now