20% Price Cut on GPT-5.6 Sol Undermines Big Law Associates’ Economics Before Harvey or Legora Need to Prove Their Worth
OpenAI reduced the pricing for the frontier model Sol by more than 20% on August 21. The cost curve works against firms reliant on senior associate hours, just before Harvey and Legora consolidate their contracts.
OpenAI announced on August 21 that it would charge $4 per million input tokens and $20 per million output tokens for GPT-5.6 Sol, its frontier model, down from $5 and $30 previously. This price cut is effective for three months and marks the second price adjustment by the provider in less than a month, following an 80% cut on Luna and a 20% cut on Terra at the end of July. The move comes four days after Anthropic informed investors that its annualized revenue run rate reached $65 billion in July, up from $47 billion in May.
The new price for Sol positions OpenAI below Anthropic's Claude Opus 5, which currently costs $5 for input and $25 for output, maintaining Sol as the benchmark frontier among major API clients. For the enterprise market purchasing tokens to automate knowledge-intensive workflows, the practical question has shifted. It is no longer about how much the model costs, but how much of a senior professional's work the model can now replace without causing material errors.
Where This Curve Meets Law Firms' Business Model
The most direct impact is on an industry that market analysis has been repeating for two years: Big Law. A typical senior associate at a New York firm costs clients between $550 and $750 per hour, of which $220 to $320 effectively compensates the professional. At Sol’s current pricing, reviewing an average complexity merger agreement consumes between $3 and $6 in tokens. This is not delicate arithmetic. An associate's hour that was relevant in 2024 has transformed by 2026 into a choice between continuing to bill the client by the hour or explaining why the hour exists.
The firms have taken notice. Clifford Chance announced staff reductions citing AI adoption in 2025. Baker McKenzie conducted a layoff round in February that internal executives attributed, among other factors, to workflow revisions with AI. Legora closed a $550 million Series D in March, and Harvey confirmed a $200 million evaluation at $11 billion, with reported usage among over 100,000 lawyers and about 1,300 organizations, many of which are part of the Am Law 100. There isn’t yet an Am Law 100 firm that has publicly declared it is in a stable growth model for its associate base.
However, there is dissonance worth noting. Danny David, head of Baker Botts, asserts that his firm will not cut associates in the wake of AI adoption. The argument is that the professional learning curve requires the same manual review that the model apparently automates, and that removing junior associates jeopardizes the partner pipeline in ten years. This is the most serious defense against the mechanical interpretation of the price drop and deserves equal weight against the optimistic view.
What the Price Drop of Sol Actually Undermines
The cut on August 21 does not change the structural argument, and should not be read as the trigger. It alters the economics of vertical buying. Vendors like Harvey and Legora needed to justify their pricing relative to the cost of running Claude or Sol directly with customized prompts. While Opus 5 was priced at $5 for input, that spread allowed for breathing space. With Sol at $4 for input, the enterprise client now compares contracts with a perceived lower price, and the value-added argument of legal SaaS shifts from "gross cost reduction" to "legal risk reduction," a much more difficult terrain to defend in RFPs with a CFO at the table.
In London, the Magic Circle firms operate with leaner staffing than the Am Law 100, and Slaughter and May and Freshfields accelerated agreements with vertical vendors since 2025. In Frankfurt and Munich, Hengeler Mueller and Gleiss Lutz have maintained investments in equivalent German tools. The partnership structure is different, so the American reading does not fully transfer. What transfers is the question: when reviewing a contract costs less than a dollar in tokens, does the firm that used to bill the associate's hour respond with hiring, or with cuts?
The Uncomfortable Evidence for Both Sides
The true trigger is not Sol’s price. It is its stability. Anthropic and OpenAI are still burning cash to sustain these curves. If the three-month promotional price ends in November and Sol goes back to $5 for input, the argument against the associate weakens. Investors in Legora and Harvey are betting on the hypothesis that the floor does not rise. It is a reasonable bet given the recent history, but it is a bet, not a theorem. A partner who decides to cut ten associates today based on a tri-monthly price table is taking HR risk within a time frame shorter than their own office lease.