A contract reviewer needs to know which clause breaks the negotiation playbook, not just whether a document looks risky. Muse Ember’s launch makes a useful question more affordable to test: can a product check individual provisions against explicit rules and assemble the exceptions for review, instead of asking for one broad assessment?
In its November 9 announcement, Meta says muse-ember-1.0 is available through the Meta Model API and Muse Code. For API prompts up to 200,000 tokens, it lists prices of $0.10 per million input tokens and $0.50 per million output tokens. Above that prompt length, both rates triple. These are token prices, not the cost of completing a contract review; retries, repeated context and human checking still affect that bill.
Consider a hypothetical small procurement team reviewing supplier agreements. Its product could send each relevant provision, the applicable playbook rule and enough surrounding text to a model, then collect proposed exceptions with source passages. Separating the checks would give the reviewer a specific disagreement to inspect: for example, whether a liability cap excludes the category the playbook requires. This is a proposed workflow, not a demonstrated Ember result.
The economic mechanism is narrower than “cheap models automate legal work.” Low token prices could make repeated, targeted checks easier to justify. If those checks preserve their evidence, the product could offer a reviewable exception list rather than another narrative the operator must reconstruct. But splitting a document also repeats context and can miss relationships between clauses; the design must establish whether that extra granularity actually helps.
A reasonable alternative is to keep the document together and use a stronger model once. Meta’s launch does not settle that choice. Its reported GDP.pdf-AA score is 22.8% at maximum reasoning effort, tied with GPT-6 Luna in the supplied comparison. That benchmark does not establish accuracy on this procurement team’s agreements.
The opportunity is therefore a more inspectable review service, not a promise of autonomous negotiation. Ember’s pricing gives builders a reason to test finer-grained checks. Whether those checks save an expert time depends on how faithfully they surface the exceptions that matter—and how much new verification work they create.
