Caesar®
A sovereign engine. It prepares the task on your infrastructure, routes it to the models you permit and checks the answer.
You decide what reaches a frontier model.
Benchmark
Caesar® against industry harnesses, industry-tuned general AI, generic LLMs and industry AI tools. Same tasks, same grading, every industry.
The comparison
Every institution picks one of these routes, often without comparing them. We run the same tasks through all five, in each industry, and score them the same way.
A sovereign engine. It prepares the task on your infrastructure, routes it to the models you permit and checks the answer.
You decide what reaches a frontier model.Agent frameworks built around frontier models for one domain: prompts, tools and retrieval wired together.
Limited to the models their vendor offers.A general model fine-tuned or packaged for a sector by its lab or a reseller.
Strong at the sector, generic in the rest.The newest frontier and open-weight models, used directly through chat or API.
Prompts go to the provider unless you host open weights yourself.Point products for one workflow in one sector, such as research, review or claims.
Built for one workflow; the rest needs other tools.Quality per dollar × reliability
Each dot is one of the newest models, scored on a public benchmark. Further right means fewer wrong answers when a model is unsure. Higher up means more quality for each dollar, ranked among the systems shown.
Higher up ranks each system’s quality per dollar among those shown: 100 is the best, 0 the lowest. Dashed lines mark the median of the models.
| Model | Quality | Cost per task | Quality per $ | Quality per $ index | Reliability proxy |
|---|---|---|---|---|---|
| Caesar® |
Quality per dollar is the quality score divided by the cost per task, counting every billable token: input, output and reasoning. At the same quality, a run that costs 10% less scores 11% higher. The chart ranks it as a percentile among the systems shown: 100 is the best, 0 the lowest, ties share a rank. The table keeps the index, where 100% is Caesar®. Compare dots within a view, not across views. Models without a published hallucination rate are in the table only.
Reliability is 100 minus the hallucination rate. Artificial Analysis counts wrong answers out of all the answers that are not correct, so a model that declines to answer scores well. It is a general proxy, not a measure of reliability in the industry you picked.
Caesar® sets the 100% mark of the quality-per-dollar index. Its reliability proxy is 88.75%. The dashed lines mark the median of the models, Caesar® left out, and split the chart into four quadrants.
Sources: Vals.ai task benchmarks (updated 7 and 8 Oct 2026) for quality and cost per task; Artificial Analysis Intelligence Index and hallucination analyses (June to September 2026) for the general reference and reliability. Gemini 4 Argon’s general-reference cost reflects a 50% launch discount; Vals.ai uses its standard price. Three hallucination rates are not from a current, matched test: Claude Fable 5.1 (launch analysis, older suite), Claude Sonnet 5.5 (pre-release deployment) and MiniMax-M3 (Artificial Analysis, June 2026).
By design
A benchmark measures answers. These are the properties that decide whether you can use those answers at all.
| Property | Caesar® | Industry harnesses | Industry-tuned general AI | Generic LLMs | Industry AI tools |
|---|---|---|---|---|---|
| You decide what reaches a frontier model. | Yes | Depends | Depends | Open-weight only | Depends |
| Choose and switch models | Yes | Partly | No | One at a time | No |
| Works across industries | Yes | One domain | One sector | Yes | One workflow |
| Every claim cited | Yes | Partly | Partly | No | Partly |
| Audit record of every call | Yes | Partly | No | No | Partly |
| Uses your own records and procedures | Yes | Yes | Partly | No | Partly |
| On-premise deployment | Yes | Depends | Depends | Open-weight only | Rarely |
Typical properties of each route as offered today. Individual products vary.
Caesar® evaluation method
The same protocol in every sector, so a score in finance means what a score in law means.
Each industry gets a task suite written with practitioners: research, document comparison, drafting and review.
Every route receives the same documents and instructions. Each runs as its vendor recommends.
Rubrics plus expert reviewers who do not know which route produced the answer. Every cited source is checked.
We publish every result, including the benchmarks Caesar® does not lead, with cost and time per task.
We divide each answer’s quality by what it costs to produce.
We measure how often an answer is wrong or cites something that does not support it.
Justinian® evidence, measured separately
Caesar® is the generation after Justinian® and takes the engine beyond law. These results are Justinian®’s own, not Caesar®’s. In law, Justinian® scores above all ten systems it was tested against, frontier models and CoCounsel included.
Source: the Justinian® benchmark, haqq.ai/justinian. Justinian® does not lead on Harvey LAB or Long Context, and we publish those results too.
Against every way to put AI to work: industry harnesses, industry-tuned general AI, generic LLMs and industry AI tools. The same protocol runs in every sector: real tasks written with practitioners, the same inputs for every route, blind grading, and every result published, including the ones Caesar® does not lead.
Justinian® is Generation 1, the legal generation: it powers HAQQ’s legal products. Caesar® is Generation 2: the same architectural foundation, extended beyond law to the records, procedures and work of other sovereign industries.
An assistant inside one vendor’s suite works with that vendor’s models and cloud. Caesar® is model-independent and sits in front of every model you approve: you decide where each request is processed, Caesar® anonymizes what must not leave, routes each task to the best approved model, cites its sources and records every call for audit. It works alongside the tools you already use.
No. Caesar® prepares the task and retrieves the relevant material before asking a model to reason, and locally hosted services handle the agreed document and context work. Each task goes to the best model you approve, so frontier quality doesn’t mean a frontier price on every call.
It retrieves the relevant material before a model reasons, compares answers across models before merging them, and cites every claim so a reviewer can check it against its source.
Yes. Every claim is cited, with references a reviewer can open to check the answer against the material.
Yes. A first deployment starts with one task, a representative set of records and a reviewer who knows the work, judged on agreed acceptance criteria: quality, time and cost.
Yes. Book a 30-minute call to see how a deployment would run for you. Then judge Caesar® on your own work: agree a task, a representative set of records and a reviewer who knows the work, and measure time and cost alongside quality.
We’ll start with your data boundary, the workflow you want to deploy and the controls your team needs to evaluate.
You don’t need a defined AI use case to start. We work through it with your team, then the architecture, infrastructure, security review and deployment path your environment requires.
Prefer email? Write to our team