Cohere

Today, Cohere is expanding its partnership with SAP to meet the increasing demand for sovereign AI technology across Europe and other key global markets. Together, we plan to make our agentic AI platform, North, available on SAP’s infrastructure to enable customers to deploy sovereign and secure AI solutions. This builds on our initial partnership which brought enterprises our frontier enterprise-grade generative and search models within SAP Business Suite, enabling agentic AI applications for customers across industries like finance, healthcare, manufacturing, and the public sector. Read the blog to learn more: https://lnkd.in/epsfnk4H

AI won’t be killing humanity. But it could put an end to boring. I sat down recently with Nick Frosst, one of the co-founders of Cohere, Canada’s premier AI company, to talk about the impact of Generative AI on work and leadership. Nick doesn’t see Large Language Models replacing valuable human work any time soon — but he does think it can remove a lot of drudgery. Other insights: 1. Study and understand your workflows, to see where AI can make a difference. “Think about your work beyond AI,” so you know where there’s leverage. 2. Challenge everyone on your team to identify work and practices that AI can do. “If AI can do X, don’t waste your time on X. That way, you can do what only you can do.” 3. Think about AI consciously — don’t let it “just happen” to your work, your teams, or your organization. 4. Leaders should teach teams to delegate to LLMs and free themselves to think strategically. 5. Treat AI as a means, not an end — a node in the system, not the system itself. 6. Let AI accelerate your culture of innovation. 7. Value those human skills that have thrived through every tech revolution: critical thinking, judgment, empathy, systems-level thinking. Best quote: “AI will not replace human capability—it will reveal it.” RBC Thought Leadership

Governments and enterprises alike increasingly recognize the importance of security and control in their AI strategy, but the path to achievement can be challenging. On our upcoming webinar, Cohere Aerospace and Defense Lead Douglas R. and Dell Technologies Enterprise Architect David Alman will share their insights on how leaders in regulated industries can develop an AI approach that adds real value. Register now to save your spot: https://lnkd.in/ejn42bMR

As part of Dell Technologies's updates to its Dell AI Factory, the Automation Platform now includes agentic AI with North, our secure agentic AI platform — enabling enterprises to quickly build scalable, secure, on-prem AI workflows. Together with Dell, we're giving businesses the control, flexibility, and performance they need to bring AI into production with confidence. Explore use cases with North within the Dell Automation Platform, like our AI co-pilot concept for finance and wealth management professionals: https://lnkd.in/gxVeUiSy

🚀 Excited to share that Cohere has contributed Shared Memory IPC Caching to vLLM! In modern LLM inference systems, data transfer between processes can become a hidden bottleneck—especially with large multimodal inputs like images and audio. Our latest benchmarks show impressive results: 📊 First-time requests: • Prefill throughput: +11.5% • Time to First Token (TTFT): -10.5% 🔥 Cached requests (with KV and image reuse): • Prefill throughput: +69.9% • TTFT: -40.5% What makes it work? Traditional IPC caching relies on strict input ordering, which breaks when schedulers reorder requests. Our Shared Memory Object Store eliminates this limitation: ✅ Single shared cache accessible by all processes ✅ No ordering assumptions needed ✅ Concurrent reads with minimal synchronization ✅ Constant memory footprint regardless of reader count The benefits scale with input size and tensor parallel degree—making it especially powerful for large multimodal workloads. Getting Started: Already available on vLLM main branch. Enable it with: `mm_processor_cache_type = "shm"` Special thanks to Donglu Wang from Cohere and the vLLM community members who made this happen: Cyrus Leung, Nick Hill, Roger Wang, and Kero Liang. 👉 Read the full technical deep-dive on the vLLM blog: https://lnkd.in/eR72Zjsr #vLLM #LLM #AI #OpenSource #PerformanceOptimization #MultimodalAI #Cohere

Do you know about Shared Memory IPC Caching? - 🧠 It keeps large multimodal inputs in shared memory - 🔀 Dramatically reduces data-transfer overhead - 🔓 Unlocks faster, more efficient LLM inference at scale Read the full article to learn more about it.   https://lnkd.in/ebB6Typ8

The path to AI adoption for enterprises can be slow and daunting without the right guidance. Watch John Watson, Group President of Bell Business, talk with Cohere Co-Founder Nick Frosst about how we are teaming up to offer end-to-end solutions and support so customers can find success with AI faster. Catch the full conversation here ➡️ https://lnkd.in/eQXKD9pB

Today’s release of the Budget 2025 by the Government of Canada is another positive step toward furthering Canada’s global AI leadership. The initiatives to facilitate government use of AI and improve capital investment for the ecosystem will play an important role in propelling Canadian AI innovation, talent, and commercialization forward. Speed is essential to ensure Canada’s competitiveness. We look forward to continuing to play a role in maintaining Canada’s competitive edge in the industry by bringing global companies and government secure, sovereign AI technology.