Cohere surfaces in Oracle AI writeups
Cohere was named in market writeups linking Oracle's embedded generative AI spend and also appeared in social mentions tied to an MXFP4 LLM training talk — multiple threads are tying Cohere into enterprise sourcing conversations. (insidermonkey.com) (x.com)
Oracle and Cohere jointly published a customer case study that says Cohere worked with Oracle Fusion Applications to build “100+” embedded generative-AI workflows across finance, HR, supply chain, sales, marketing and service. (cohere.com/customer-stories/oracle, ) Oracle’s formal announcement of its OCI Generative AI initiative on June 13, 2023 explicitly named Cohere as a foundation-model partner for the service that spans applications to infrastructure. (oracle.com/news/announcement/oracle-to-deliver-powerful-and-secure-generative-ai-service-for-business-2023-06-13, ) Oracle’s engineering blog and docs list Cohere Command (Command‑A and Rerank) models among the foundational models available on the OCI Generative AI service, and Oracle’s technical case study states Cohere trains and deploys models on OCI (case study published May 20, 2025). (blogs.oracle.com/ai-and-datascience/cohere-command-a-rerank-oci-gen-ai, ) (oracle.com/cloud/technical-case-studies/cohere/, ) Market coverage of Oracle’s recent results and strategy tied OCI infrastructure demand to application-embedded AI, with analyst writeups on Q3 FY2026 noting contracted GPU capacity build and backlog expansion tied to embedded/agentic AI features. (futurumgroup.com/insights/oracle-q3-fy-2026-earnings-driven-by-oci-ai-infrastructure-demand/, ) The AISTATS/ML research “Training LLMs with MXFP4” (Amazon Science) presents a near‑lossless MXFP4 training recipe that trains GPT‑scale models up to 6.7B parameters and reports >1.3× speedup vs FP8 and >1.7× vs BF16 for backpropagation. (proceedings.mlr.press/v258/tseng25a.html, ) A public X post cited in the card (x.com) is among social threads that linked the MXFP4 training talk to enterprise LLM vendors; the original post and its replies include mentions of model providers and deployment tradeoffs in threads tied to the MXFP4 paper. (x.com/i/status/2034031630399013274, ) Oracle’s OCI Generative AI pricing and docs describe “AI units” and two dedicated AI cluster types (hosting and fine‑tuning), specifying how customers assign AI units to hosting clusters based on model and expected call volume. (oracle.com/artificial-intelligence/generative-ai/generative-ai-service/pricing/, )
