AI21 repositions Maestro from orchestration to agent cost optimization
AI21 now sells Maestro as an optimization framework that cuts what production AI agents cost, adding per-team, per-repo and per-agent spend attribution. Jamba token rates unchanged.
Maestro sold as an AI planning and orchestration system that plans, validates and self-corrects multi-step tasks, governed by a budget parameter trading speed against cost and reliability.
Maestro sold as an optimization framework for real-world AI agents that routes, compresses and decomposes every request in flight with no code change, maps the cost / accuracy / latency frontier, and attributes spend to the team, repo and agent that caused it. Still contact-sales only, no public price.
AI21 Labs rebuilt the Maestro product page around cost, not capability. The headline moved from planning and orchestration to “Optimize cost, accuracy and latency in real-world AI agents”, and the lead claim is now financial: “Most agent budgets are over half recoverable waste.” Maestro promises to route, compress and decompose agent requests in flight without code changes, and to attribute spend down to the team, repo and agent that caused it — FinOps-style allocation applied to agent runs rather than cloud resources.
The pricing mechanics behind it did not change: Maestro carries no published price on either the product page or the pricing page and remains contact-sales only, with the budget concept surviving as “automatic budget & compute scaling” that keeps parallel execution paths inside a stated cost and latency budget. The self-serve sheet is also unchanged — Jamba Mini at /bin/bash.2 in / /bin/bash.4 out and Jamba Large at in / out per 1M tokens, a 0 trial credit, Pay As You Go with unlimited seats, and a quoted Custom Plan.
One inconsistency worth noting for buyers: the marketing pricing page advertises “0 credits for 7 days”, while AI21’s docs pricing page describes the same 0 credit as “good for three months”.
The Maestro page drops the 'AI planning & orchestration system' framing for 'an optimization framework for real-world AI agents' that 'cuts what they cost — without touching what they deliver', adding per-team / per-repo / per-agent spend attribution and a cost/accuracy/latency frontier. Jamba token rates are unchanged (Mini $0.2/$0.4, Large $2/$8 per 1M) and Maestro stays quote-only.