{"city":"San Francisco","name":"Together AI","slug":"together-ai","facts":[{"date":"2026-09-10","type":"deployment","value":"Preemptible GPU clusters deployed for internal research, development, testing, and non-critical production workloads","source":"Together AI"},{"date":"2026-09-10","type":"product_launch","value":"Automated reclamation and refill for preemptible GPU nodes","source":"Together AI"},{"date":"2026-09-10","type":"product_launch","value":"Preemptible GPU compute priced at 50% of the on-demand rate","source":"Together AI"},{"date":"2026-09-10","type":"product_launch","value":"Public preview of preemptible compute for Together GPU Clusters","source":"Together AI"},{"date":"2026-09-10","type":"product_launch","value":"Preemptible capacity controls exposed through the cloud console, CLI, and API","source":"Together AI"},{"date":"2026-09-02","type":"partnership","value":"Inference Exchange collaboration with Equinix and NVIDIA to optimize AI inference deployment","source":"Equinix"}],"peers":[],"stage":null,"state":"California","aliases":["Together","Together Computer"],"country":"United States","domains":[{"why":"Together AI runs an AI cloud offering inference, hosted model APIs, fine-tuning and GPU clusters to enterprises and developers.","list":null,"rank":null,"tier":"watchlist","layer":null,"total":80,"domain":"AI Core","scores":{"gate_ne":["firepower"],"leverage":8,"position":6,"progress":10,"firepower":8,"commercial":8,"deployment":10,"production":10,"technology":6},"edition":null,"segment":null}],"founded":null,"signals":[{"date":"2026-10-07","header":"Together AI, Modal, and Base10 are each operating open-model businesses at billion-dollar run rates.","outlet":"Peter H. Diamandis"},{"date":"2026-09-10","header":"Together AI’s optimized Vera Rubin NVFP4 GEMM exceeded 22 PFLOPS.","outlet":"Together AI"},{"date":"2026-09-10","header":"Together AI added NVIDIA Vera Rubin NVL72 support to ThunderKittens GEMM kernels.","outlet":"Together AI"},{"date":"2026-09-10","header":"Together AI measured large GEMM speedups from Vera Rubin’s early A-release instruction.","outlet":"Together AI"},{"date":"2026-09-10","header":"Together AI is hiring for its kernels and performance teams.","outlet":"Together AI"},{"date":"2026-09-10","header":"Together AI’s deeper Vera Rubin pipelines reached 22,239 TFLOPS on a 16k-square NVFP4 GEMM.","outlet":"Together AI"}],"website":"together.ai","category":"AI Infra · Compute software","categories":[{"name":"Neoclouds & GPU clouds","domain":"AI Core","parent":"AI Infra · Compute demand & operators"}],"last_round":null,"updated_at":"2026-10-25T23:19:36.455605+00:00","facts_total":135,"positioning":"Open-model inference cloud and GPU-cluster renter that crossed $1B run rate; building its own data centers (Maryland, Memphis) rather than only reselling CoreWeave and Lambda capacity; the reference name for open-weight inference economics.","sub_category":"Inference & serving clouds","signals_total":299,"url":"https://www.frontiertechintelligence.com/companies/together-ai","title":"Together AI · ai core, San Francisco, California, United States · scores, facts and signals · Frontier Tech Intelligence","description":"Together AI runs an AI cloud offering inference, hosted model APIs, fine-tuning and GPU clusters to enterprises and developers. AI Core watchlist (score 80) 299 signals and 135 facts with sources.","publisher":"Frontier Tech Intelligence (Future Frontiers, a trade name of Essential Cognition Inc.)","note":"Assessment by Frontier Tech Intelligence; facts and signals carry their sources."}