
Perplexity unveils Computer: a 19-model orchestration platform
Context and Chronology
Perplexity announced a new product named Computer that runs in the cloud and autonomously breaks complex projects into subtasks, delegating work across many backend models. The launch package is first available to subscribers of the company\'s premium tier, priced at $200 per month, positioning the feature as a monetized capability before broad enterprise rollout. CEO Aravind Srinivas framed the product as an orchestration layer that treats models as interchangeable tools rather than monolithic endpoints; Mr. Srinivas emphasized model specialization as the commercial inflection point for Perplexity\'s strategy.
Technically, the system routes tasks to specialist models: a reasoning and orchestration kernel runs on Claude Opus 4.6, deep research tasks fall to Gemini, image and video assets are handled by dedicated generators, and long-context retrieval uses GPT-5.2. The roster is intentionally fluid — new models may be added as they show domain strength, and users can override automated routing to assign roles manually. That architecture converts model heterogeneity into a product feature instead of a compatibility headache.
Perplexity presented enterprise usage telemetry showing a rapid move away from single-model dominance: early last year, most tasks clustered on two models, but by year\'s end no single model exceeded a quarter of usage across enterprise workloads. Executives reported that frontier models surfaced at a cadence measured in days, creating a fast-changing supplier base and making single-vendor lock-in less attractive to multi-disciplinary teams. Those trends gave the company confidence that an orchestration layer would capture disproportionate value from mixed-model workflows.
Operationally, Perplexity has formalized hosting on Microsoft Azure, a move the company portrays as both an operational contingency and a strategic alignment that secures scalable compute, data pipelines and potential go-to-market support. That arrangement improves short-term continuity for Computer but also ties the orchestration layer to a hyperscaler relationship with commercial and competitive implications: Microsoft gains a high-profile tenant to bolster Azure\'s credentials, while Perplexity benefits from predictable infrastructure and potential integration pathways into Microsoft\'s enterprise tooling.
The Azure hosting decision also surfaced commercial tension with other hyperscalers — most notably a public dispute with Amazon over infrastructure terms — underscoring the reputational and contractual risks that arise when startups rely heavily on single cloud partners. Negotiations around egress fees, privileged-access clauses, compute allocations and exit rights are now receiving heightened scrutiny from procurement teams and regulators, because such terms can materially affect portability and neutrality across clouds.
Security and operational choices are central to the product\'s differentiation. Computer executes inside a cloud sandbox, a deliberate contrast with local agents that require filesystem access and can interact with on-device APIs. That decision follows public incidents involving autonomous local agents and feeds an explicit trust argument: contain failure modes in the cloud instead of exposing enterprise endpoints to uncontrolled agent behavior. Perplexity also stresses accessibility from lightweight clients — phone, chat, or browser — to lower activation friction compared with terminal-based tools.
Beyond the consumer face, Perplexity disclosed that its search API is already running in production inside several major technology companies, and executives described a feedback loop where search ranking and model consumption reinforce each other. The company reported faster revenue growth than user growth over the past year, signaling improved monetization per account and a potential enterprise runway. Legal exposure continues to shadow expansion: ongoing copyright disputes and commercial frictions with publishers remain unresolved and could affect data sourcing and product features as the company scales.
Taken together, the product launch and the Azure hosting arrangement illustrate a strategic trade-off: Computer aims to decouple value from individual model makers by routing across specialists, yet its dependence on a single hyperscaler for critical infrastructure creates a second axis of vendor leverage that procurement, rivals and regulators are likely to watch closely.
Read Our Expert Analysis
Create an account or login for free to unlock our expert analysis and key takeaways for this development.
By continuing, you agree to receive marketing communications and our weekly newsletter. You can opt-out at any time.
Recommended for you

AT&T Rewrites Model Orchestration, Cuts Costs by 90%
AT&T rearchitected its model orchestration to route work across many smaller models, achieving up to 90% cost savings while handling scale of roughly 8 billion tokens daily. The new stack, built on LangChain and deployed with Microsoft Azure , has been rolled out to over 100,000 employees and materially shortened developer cycle times.

Perplexity Turns to Microsoft Azure for AI Hosting as Tensions with Amazon Flare
Perplexity has formalized a hosting arrangement with Microsoft Azure to support its AI services while navigating a public fracas with Amazon’s cloud unit. The move underscores a broader industry trend—hyperscalers pairing capital, privileged hosting and commercial ties to shape access to leading models, raising scrutiny over lock‑in and interoperability.

Multiverse Computing bets on compressed models for on-device AI
Multiverse is shipping a mobile-first app and a self-serve API that expose its compressed models, pitching privacy, resilience, and lower inference cost. The move accelerates edge deployment while forcing enterprises to re-evaluate cloud compute commitments and procurement timelines.

OpenAI’s GPT-5.4 Brings Native Computer Control and Deep Spreadsheet Integration
OpenAI released GPT-5.4 with built-in computer-control and spreadsheet plugins, boosting long-context capability and cutting token costs in targeted workloads. Pricing and a new long-context surcharge reshape enterprise automation economics and developer trade-offs.
Cursor: Composer 2 Built on Moonshot’s Kimi Exposes Western open-model gap
Public interception showed Composer 2 runs on Kimi K2.5 , forcing urgent vendor disclosure and enterprise provenance checks. The incident highlights a persistent Western open-weight shortfall and elevates procurement, licensing, and geopolitical risk for AI products.

NVIDIA unveils Nemotron 3 Super for enterprise agents
NVIDIA released Nemotron 3 Super, a reasoning‑first model aimed at sustained, multi‑step enterprise agents and published with open weights, datasets and recipes to enable on‑prem deployment and fine‑tuning. Public reports differ on headline parameters (the company and some outlets cite ~120B while other engineering notes and press accounts describe ~128B), but all sources confirm a runtime sparsity mode (reported as ~12B active parameters) plus a wider program and hardware roadmap—NemoClaw, NVL72/Rubin racks and privileged partner access—that together reshape procurement and vendor leverage for enterprise agent stacks.
Gimlet Labs Raises $80M to Orchestrate Multi‑Silicon Inference
Gimlet Labs closed an $80M Series A led by Menlo Ventures to commercialize a multi‑silicon inference cloud that shards agentic workloads across heterogeneous hardware. The raise and product launch sit inside a broader wave of infrastructure bets — from edge runtimes to stateful AI platforms — that collectively signal software orchestration is becoming the primary lever for lowering inference cost and shaping procurement.
OpenAI accelerates theoretical-physics calculations with model collaboration
OpenAI -backed models helped researchers solve complex gluon calculations, producing two preprints in early 2026 and compressing timelines from months to weeks. Company-published usage statistics and cross‑vendor demonstrations suggest this episode is part of a broader move toward agentic, model-in-the-loop scientific workflows — but widespread adoption depends on urgent investment in provenance, formal verification and new institutional practices.