HAQQ Academy - free, and open to anyone
Get certified in legal AI. Free.
33 chapters on doing legal work with AI, written by the team building Justinian - the legal AI engine whose answers come with citations you can verify. Pass the 33 chapter tests and the certificate at the end is yours.
- No account, no card, no credits
- Self-paced - most people take a week
- Shareable certificate at the end
HAQQ Labs
How the legal AI field actually performs
We publish the research behind Justinian: independent benchmark results on reliability and cost, the frameworks the field uses to evaluate legal AI, and the experiments we ship in the open.
Task Benchmark
Legal Research Bench
Vals.ai, updated October 24, 2026. US legal research tasks.
Reference
Justinian®
The reference the source page sets its scale on. It takes the top rank by definition, so its position is not a measurement.
Frontier Labs
- Anthropic
- OpenAI
- xAI
- Meta
Open-weight
- DeepSeek
- Qwen
- Moonshot AI
- Z.ai
- MiniMax
- Xiaomi
- Mistral AI
- Claude Opus 5.5: 41.0% / 2.0
- Claude Fable 5.1: 27.4% / 2.4
- Claude Sonnet 5.5: 53.0% / 3.2
- GPT-6 Astra: 49.0% / 3.7
- Gemini 4 Argon: 85.0% / 8.5
- Grok 4.7: 71.0% / 9.6
- GPT-6.1 Sol: 46.0% / 11.4
- MiniMax-M3: 83.9% / 87.7
- Justinian®: 88.8% / 100
How to read this chart
Task benchmark: Legal Research Bench on Vals.ai, updated October 24, 2026. Scope: US legal research tasks, not every jurisdiction or compliance workflow.
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. Justinian® is the source page's reference system, so it takes the top rank by definition. This is not a new measurement.
Not plotted, no published reliability rate: MiMo V2.6 Flash, GLM 5.3 Flash, MiMo V2.6 Pro, DeepSeek V4.1 Flash, Muse Spark 1.3 Max, Claude Opus 5, GLM 5.3, Qwen 3.8 Max, Kimi K3, Claude Haiku 5.5, Mistral Large 4.
How you get there, in three steps
Free, self-paced, and open without an account. The only thing it asks of you is that you try each chapter on real work.
Take the course, and use it on real work
33 chapters, from what a large language model actually is to running a matter end to end. Read a chapter, then open HAQQ in the next tab and run it on a file you are already working on. Nothing here is theory you cannot try the same afternoon.
Pass the test at the end of every chapter
33 of the 33 chapters end in a test, so you find out what you actually retained instead of assuming. The rest of the Academy is open while you work: 631 prompts, 738 clause examples across 44 categories, 422 AI skills, 70 document templates and 22 free tools. No account, no credits.
Get certified
Pass the test at the end of every chapter that has one and you have completed HAQQ's legal AI course - a certificate for your LinkedIn profile, your CV and your feed that says what you can do on Monday morning, not what you sat through. It is free, like the rest of it.
Four thousand years, in twenty-nine chapters
The course reads law as one long protocol upgrade - rules cut into stone, then copied by hand, then sold by the hour, then run by machines. Six of the chapters it passes through.
Certification program
Earn it one track at a time.
The same 33 chapters, cut into 3 certifications you can earn one at a time. Finish them all and the full certificate above is still waiting at the end.
- Level ILegal FoundationsHow law actually works, and how the business of law actually pays.Chapters 1 to 12 · 12 Quizzes
- Level IILegal AI PractitionerDoing real legal work with AI, without giving away privilege.Chapters 13 to 21 · 9 Quizzes
- Level IIILegal AI StrategistChoosing the stack, delegating the work, and proving it paid.Chapters 22 to 33 · 12 Quizzes
- Every levelFull course certificatePass all 33 chapter tests and claim a shareable certificate for LinkedIn and your CV.See your certificate
Every page here is free and open without an account. The product it teaches is not, that is the only difference.
HAQQ Academy - free, and open to anyone
Everything on this site
The course, the certifications, the prompts and the research behind all of it.






