Perplexity has cut its annual enterprise seat prices by roughly 15%, dropping Enterprise Pro from $40 to $34 per seat and Enterprise Max from $325 to $271. The change lands alongside a broader restructuring of the developer platform, where the minimum credit purchase for new API accounts fell from $50 to $10. This reduction in entry barriers is significant for teams testing the infrastructure before committing to production workloads. The company also retired the Sonar Chat Completions endpoint on September 27, forcing all new integrations to migrate to the Agent API. Existing synchronous requests continue working through internal reformulation, but asynchronous calls are no longer supported and must move to background mode.
Enterprise Seat Adjustments
The new pricing structure applies specifically to annual billing cycles, with monthly rates remaining unchanged at higher points. Enterprise Pro now includes 8,500 bonus credits and 500 monthly Computer credits per seat, while Enterprise Max provides 40,000 bonus credits and 15,000 monthly Computer credits. These tiers add SSO, user permission management, and dedicated support to the base Pro features. Larger deployments requiring advanced security or governance still receive custom quotes under contract. The price drop aligns with Perplexity’s push to consolidate fragmented AI adoption onto a single governed platform, reducing the cost of scaling across large organizations.
| Plan | Annual Price (Old) | Annual Price (New) | Monthly Credits |
|---|---|---|---|
| Enterprise Pro | $40/seat | $34/seat | 500 |
| Enterprise Max | $325/seat | $271/seat | 15,000 |
The shift reflects a strategic move away from advertising toward usage-based subscriptions, a model CEO Aravind Srinivas describes as the right long-term business approach for AI. By lowering the per-seat cost, Perplexity aims to make its "AI operating layer" more accessible to mid-sized enterprises that previously found the entry price prohibitive. This adjustment positions the product against competitors like Glean, which tailors pricing based on organization size and specific search requirements.
Developer Platform Changes
On the API side, the standard web_search invocation cost dropped from $0.005 to $0.0025 per call, halving the expense for agent workflows that rely heavily on retrieval. Fast Search remains at $0.001 per invocation, while image_search costs $0.0025 per successful call. The sandbox environment, used for code execution, bills at $0.03 per session covering up to 20 minutes of active use. These rates apply to the Agent API, which routes requests to models from OpenAI, Anthropic, Google, and others. The Orchestrator API, currently in private preview, bills per token at each model’s published rates without additional request fees.
Usage tiers govern rate limits based on cumulative lifetime spend rather than current balance. Tier 0 allows 1 query per second, while Tier 5, unlocked at $5,000 in purchases, permits 33 queries per second. Once an account reaches a tier, it keeps that status permanently even if spending slows down. The platform can also be purchased through AWS Marketplace as a SaaS listing starting at $1,000 in credits. This route requires post-purchase registration to provision the user and generate an initial key.
The retirement of plain sonar-reasoning in December 2025 left sonar-reasoning-pro as the only reasoning-tier model, a change that caught some developers off guard. Any code still calling the old name targets a model that no longer exists. With the minimum purchase now at $10, new users can test authentication and basic functionality without a significant upfront commitment. However, the complexity of managing multiple tool invocations and token counts means that a simple lookup can quickly accumulate costs if not monitored closely. Does the lower entry price offset the variable compute costs hidden within multi-model orchestration?
