Our recommendation: choose the advantage your practice needs
For a Swiss firm, the best legal AI is the one that improves the work you actually deliver. Start with the bottleneck: finding authority, using the firm's knowledge, reviewing a document, or turning instructions into a coherent matter and draft. Then compare the relevant products on the same completed matter.
Shortlist Silex and Omnilex for a Swiss research-and-workspace evaluation; Lawsearch AI for Weblaw source access; Swiss-Noxtua for the relevant legalis commentary; Jurata, Libra and CASUS when Word or publisher-backed research shapes the decision. Evaluate Whisperit when spoken instructions, matter context and document preparation are the daily problem. These are our recommendations about fit, not measured rankings. [1] [2] [3] [4] [5] [6] [7] [8]
The opportunity is larger than a quicker answer. A good tool can help a lawyer return to a matter with its reasoning intact, give a colleague a clear starting point and spend more attention on the decision that deserves it. That promise should survive a real evaluation.
Decision diagram
Three kinds of knowledge to check separately
Public law
Legislation and court decisions: scope, date and full text.
Licensed commentary
Publisher collections: entitlement, edition and citation depth.
Your firm’s knowledge
Matter files and precedents: permissions and traceability.
A Swiss-law answer and a matter-grounded answer require different evidence. Confirm the exact collections, licence and access rights in the plan you are evaluating.
The Swiss market is several buying decisions
“Swiss legal AI” can describe a Swiss company, Swiss legal content, Swiss storage or Swiss model processing. Those properties can occur separately. A firm should write down which ones it needs before comparing logos.
The following map groups products by the reason we would evaluate them. Each product can extend beyond its listed entry point; the rows are not an exhaustive feature inventory.
Research and source collections
| Product | Why we would evaluate it | What could change the recommendation |
|---|---|---|
| Silex | Swiss research with inspectable reasoning | The exact cantonal coverage, historical versions and workflow result |
| Omnilex | Swiss research within a broader software workspace | Provenance of commentary and actual case collaboration |
| Lawsearch AI | Research using public sources and Editions Weblaw full texts | Whether those publications match the practice's needs |
| Swiss-Noxtua | Research requiring relevant legalis commentary | The licensed collection and selected processing configuration |
Sources: the vendors' current research and product descriptions. Corpus counts and promotional time savings are not used as comparative accuracy evidence. [9] [2] [3] [4]
Documents and daily production
| Product | Why we would evaluate it | What to inspect in the evaluation |
|---|---|---|
| Jurata | Swiss research and drafting with a Word add-in | Source validation, Word edits and the applicable plan |
| Libra | Stämpfli-linked research and Microsoft workflow | Publisher module, Word tier and actual integration scope |
| CASUS | Native Word review, proofreading and document benchmarking | Usable changes, party-specific risk instructions and broader matter work |
| Whisperit | Instructions, matter context and reviewed document preparation | Capture quality, factual consistency, source inspection and export |
Sources: official product and packaging pages. Word export and an interactive Word add-in are different capabilities. Whisperit's published integration material establishes export; it does not establish the same in-editor workflow as these competitors. [10] [11] [7] [12]
Commentary is not one interchangeable feature
A tool can cite official legislation and decisions without granting access to the commentary your firm relies on. Ask for a named work, edition, author, passage and licence entitlement. A headline promising a large corpus cannot answer that question.
Four examples make the distinction concrete. Swiss-Noxtua documents legalis research. Libra offers a Stämpfli content connection and a separately priced Legal Premium module. Jurata describes Orell Füssli commentaries in a pilot, with availability subject to change. Omnilex describes its own AI-assisted commentary; this should not be assumed equivalent to a licensed publisher collection. [4] [13] [10] [2]
Lawsearch's access to Editions Weblaw is another specific collection, not a claim to include every Swiss publisher. Keep your existing research subscriptions until the replacement has demonstrated the sources you actually need. [3]
Swisslex belongs in this purchasing discussion as a source library. Its public overview describes journals, commentary and case law. We have not audited its current AI offering or established professional subscription prices here; this guide makes no claim that Swisslex lacks AI. The question is whether a proposed replacement preserves the authority and full-text access in the firm's existing licence. [14]
Where the other Swiss players fit
DeepJudge deserves attention when the firm's accumulated knowledge is the asset: previous matters, internal documents and information spread across repositories. That is a different entry point from buying access to a particular Swiss commentary. A larger knowledge program should test repository integration and access control as well as search relevance. [15]
OpenCaseLaw supplies an open, machine-readable record of published Swiss law and an AI connection. Weblaw tested a model combined with OpenCaseLaw. Evaluating that combination requires assessing the model and its data path too; the source connector alone does not establish the security or capabilities of a complete legal workspace. [16] [17]
Unplex participated in Weblaw's June challenge. Its current homepage positions the workspace around Swiss wealth-management work. Confirm the present legal research package before treating historical challenge participation as proof of the product you can purchase today. [18]
These adjacent choices matter because a firm may sensibly use a source library, an internal search system and a production workspace together. One subscription replacing everything is an attractive hypothesis, not a starting assumption.
What Weblaw adds—and what it cannot settle
Weblaw's 17 June 2026 challenge brought seven tools to four Swiss cases. Its public material is useful evidence about the kinds of errors worth testing. The available result matrices have anonymous columns; they do not establish a named provider winner. Weblaw also operates participant Lawsearch AI, so this is organizer-published evidence with a commercial interest, rather than an independent benchmark. [17]
The public case-two matrix shows six of seven columns citing the expected leading decision. The case-three matrix shows all seven finding the expected decision, with six fully and one partially matching the expected answer. Those observations concern those tasks; they are not general product accuracy rates. [19] [20]
Read the detailed evidence analysis and download its anonymous observations. Use the challenge to sharpen an evaluation, rather than to award a winner it does not identify.
Processing location needs its own comparison
Published Swiss storage does not establish Swiss-only inference. Omnilex and Jurata describe Swiss storage with Swiss/EU processing. CASUS specifies Swiss storage and inference in Switzerland, Belgium and Sweden. Libra describes German storage and EEA processing. These are vendor statements about their current offerings, not our audit of a tenant. [21] [22] [23] [6]
Silex's architecture page describes Swiss storage and inference, while its privacy notice permits Switzerland and/or EU hosting and transfers. Reconcile the scope in the proposed agreement. Whisperit documents Swiss Only, EU and Global configurations; select the configuration rather than applying one description to every plan. Swiss-Noxtua likewise distinguishes processing options. [24] [25] [26] [27]
If a client policy requires all relevant processing to remain in Switzerland, make that a documented procurement condition. It is not necessary to claim that every other arrangement is legally impermissible. A product's research performance cannot substitute for the selected data-flow evidence.
Public prices: keep the unit visible
These are selected annual offers displayed on the checked pages, not matched all-in proposals or a price ranking.
| Vendor and selected plan | Displayed annual fee, CHF | Important scope difference |
|---|---|---|
| Silex Solo / Team | 1,440 / 5,280 | Solo one seat; Team up to five; finite agent credits |
| Omnilex Individual / Team | 2,000 / 1,600 per user | Team starts at two users; annual per-user billing |
| Jurata Pro | 1,920 per user | Derived from 160/month billed annually; Word included |
| Libra Professional | 2,080 per user | Promotion through 31 December 2026; content add-on separate |
| CASUS Standard | 1,440 per user | Advanced governance and rollout quoted separately |
Silex, Omnilex, Jurata, Libra and CASUS explicitly label these prices as excluding VAT. Confirm the current offer before purchasing. Sources: [28] [21] [10] [29] [30]
Whisperit, Lawsearch and Swiss-Noxtua should enter the commercial comparison with a current scoped offer, not an invented comparable rate. Include source modules, seats, usage, support, retained subscriptions and renewal conditions. A flat team fee is different from a per-user tariff.
Build a shortlist you can finish evaluating
Choose two or three products around one recurring job. Give them approved material, identical instructions and the same expected deliverable. Record the plan, sources, model configuration and date. Test historical law, a contrary decision and a deliberately missing fact. Then ask a second lawyer to verify and continue the work.
Measure time to a reviewed result, including corrections and handover. Obtain comparable commercial offers with seats, usage limits, source modules, support and retained subscriptions. Published entry prices are useful starting points; they are not all-in costs for the same system.
The Swiss evaluation scorecard provides a blank worksheet. A small, well-designed trial can produce a decision more valuable than a sweeping ranking: “For our matters, with these controls and this review process, this tool makes the work better.”
Frequently asked questions
Is Whisperit the best Swiss legal AI?
We have not established that claim. Its proposed advantage is the connected daily workflow. A competitor may be preferable for licensed commentary, in-editor review or a substantial internal knowledge program. Demonstrate the actual workflow before deciding.
Should Silex and Omnilex be in the original top five?
They belong in a focused Swiss shortlist. The original five cover broader strategic buying decisions. This supplementary collection makes local overlap explicit without pretending either list is a market-share ranking.
Can Weblaw's results decide between Silex and Omnilex?
The public anonymous matrices cannot support that attribution. Our Silex vs Omnilex comparison instead uses current product documentation and a common trial design. No provider-level challenge score is invented.
Romain de Wolff
