Home / Compare / Mars vs Torzon
Head to head
Mars vs Torzon (2026): A Graded Project Against an Unrated One
Most comparisons on this site put two completed scorecards beside each other. This one cannot, and the reason it cannot is the most useful thing on the page. Torzon has been through the rubric and carries a Caution grade. Mars has not, and printing a number for it would be invention rather than journalism.
What Unrated is, and what it is not
A reader arriving at this page probably wants to be told which of the two is better, and the honest response is that only one of them has been assessed. Unrated on this site means the rubric could not be completed against evidence we were able to check ourselves. It is not a euphemism for suspicion, it is not a soft warning, and it is not a placeholder we intend to quietly upgrade once sentiment turns favourable. It is a statement about our own coverage, and the failure it describes is ours rather than the project's.
Treating an absence of assessment as though it were a poor assessment would be just as dishonest as inventing a favourable one. Both errors put a number where evidence is missing. What a reader can legitimately take from Unrated is narrow and specific: this desk cannot help you decide about Mars yet, so any decision you make about it is being made on someone else's evidence, and you should be able to say whose.
Side by side, where a comparison is possible at all
| Dimension | Torzon | Mars | Status |
|---|---|---|---|
| Uptime feel | Reachable across most checks; brief gaps clustered around rotations | Not sampled on a schedule sufficient to characterise | Torzon assessed |
| Encryption and PGP | Enforced by default; key handling sane at the visible layer | Signing practice not independently confirmed by this desk | Torzon assessed |
| Interface | Fast and legible at the strictest browser setting | Observed only in passing; no basis for a judgement | Torzon assessed |
| Support responsiveness | The documented weak point; queues stretch around rotation windows | No corroborated reports we could verify | Torzon assessed |
| Mirror rotation discipline | Frequent rotation; each change reopens a phishing window | Cadence unknown to us | Torzon assessed |
| Dispute handling | Inconsistent outcomes; drift on longer cases | Unassessed | Torzon assessed |
| Community trust | Long presence, mixed sentiment, no disqualifying event on record | Sentiment exists; verification of it does not | Torzon assessed |
| Overall | Caution | Unrated | Not comparable |
The right-hand column is not padding. A table with one side empty is an uncomfortable thing to publish, and it is a more accurate artefact than the alternative, which would be a tidy grid of confident-looking entries assembled from forum posts and inference.
What is actually known about Torzon
Torzon's completed scorecard is documented in full on its own review page, and the summary version is that its infrastructure is better than its service. The address answers reliably outside rotation windows, encryption is enforced rather than optional, and the interface holds up under the browser configuration a careful reader should be using. The Caution grade is driven almost entirely by what happens after something goes wrong: support that stretches, disputes that lose momentum, and a rotation cadence frequent enough that re-verification cannot be treated as an occasional chore.
That profile has a specific consequence for anyone comparing it against an unknown. Torzon's weaknesses are legible. You can plan around a slow dispute queue because you know it is there. Planning around an unassessed project means planning around a shape you have not seen, and the fact that no weaknesses appear in the Mars column above reflects the absence of an investigation rather than the absence of problems.
Why Mars has not been graded
The rubric requires evidence this desk gathered or independently corroborated, and for Mars that work has not been done to the standard the other reviews on this site were held to. There is no dramatic explanation. Assessing a project properly means repeated observation over time, confirmation of signing practice against a source we did not receive from the project itself, and enough corroborated accounts of dispute and support behaviour to distinguish a pattern from an anecdote. Partial versions of those inputs exist. Completed versions do not.
Publishing anyway would be easy and it would be worthless, because a fabricated assessment is indistinguishable, to the reader, from a real one. The whole value of a grade on this site rests on the grades being refusable. If Unrated were never printed, Recommended would mean nothing, and a reader would have no way to tell an evaluated project from an unevaluated one dressed up to look the same.
What would change the status
Three inputs, all of which take time rather than luck. Repeated reachability observation across a long enough window that rotation behaviour becomes visible as a pattern. Independent confirmation that published statements verify against a key obtained from somewhere other than the project itself. And corroborated accounts of how disputes actually resolve, from enough separate sources that a single loud voice cannot drive the finding. When those are in hand, Mars gets a scorecard like any other project here, favourable or not.
A leaderboard with an empty column is embarrassing to publish and honest to read. The reverse arrangement is far more common.
How to act sensibly around an unrated project
Unrated is not an instruction to walk away, and readers who treat it that way are overreading it in the opposite direction from the readers who ignore it. What it does establish is that the burden of evidence has moved onto you, and the practical question becomes whether you are equipped to carry it. Three things follow from that, and they apply to any unassessed project rather than to Mars specifically.
Know whose evidence you are actually relying on
If you proceed without an assessment from this desk, you are proceeding on somebody else's. That is entirely legitimate, provided you can name the source and say what work it did. "Several people say it is fine" is not a source; it is an aggregate of unexamined impressions, and aggregating impressions does not convert them into evidence. A source worth relying on is one whose method you could describe to somebody else and whose incentives you have thought about, particularly the incentive to be listed and paid for it.
Match your exposure to your confidence, not your curiosity
The unmapped-risk problem has a straightforward mitigation: keep the amount at stake within what an unknown failure mode could take without consequence. This is duller advice than a grade would be and it is more robust, because it does not require knowing which failure mode is the relevant one. Curiosity about an unfamiliar project is a fine reason to look. It is a poor reason to commit anything you would mind losing.
Build your own record before you trust the project
Repeated small interactions over time produce exactly the kind of evidence a scorecard is made from — how quickly a message is answered, whether an address changes without warning, whether the behaviour you saw last month is the behaviour you see this month. A reader who accumulates that deliberately, and who is willing to revise downward on what they find, ends up better informed about their own experience than any external grade could make them. That is not a substitute for this desk doing its work. It is what a careful person does in the meantime.
What the asymmetry costs, and why we keep it
There is a real price to publishing a comparison with one side empty, and it is worth acknowledging rather than glossing. A reader who wanted a verdict leaves without one. A page that could have been decisive is instead careful, and careful pages are less satisfying to read and less useful in the moment somebody actually wants help. This desk accepts that cost because the alternative degrades everything else on the site.
The mechanism is straightforward. Every grade published here draws its meaning from the grades that were withheld. If the rubric could be completed against inference and sentiment whenever real evidence was inconvenient to gather, then a completed scorecard would no longer tell a reader that the work was done — it would only tell them that a page was filled. At that point the difference between this journal and the directories that will list anything for a fee collapses, and the collapse is invisible from the outside, because both kinds of page look identical.
Keeping the column empty is the only version of this comparison that remains true if the reader checks. That is the standard we would want applied to us, and it is the reason a page that admits what it does not know is more useful than a page that appears to know everything.
Evaluating an unrated project yourself: a working checklist
If you decide to proceed around an unrated project, you are effectively doing the work this desk has not finished, so it helps to know what that work actually consists of. The rubric is not a secret; it is a discipline, and the checklist below is the honest version of what "verify it yourself" means when nobody hands you a grade. None of it requires special tools. All of it requires patience, which is precisely the resource a phishing operator is betting you will not spend.
- Establish the address before anything else
- Obtain the current address from a source that signs what it publishes, verify that signing key against something other than the project's own front page, and confirm the same address appears identically in a second independent place. Read all 56 characters. This single step prevents the failure mode that costs readers more than every other risk combined, and it comes before any judgement about the market's quality.
- Sample reachability over time, not once
- A single successful connection tells you almost nothing. Observing that a project answers across many checks over days, with only the circuit-dependent slowness normal to Tor, is what distinguishes a stable service from one you happened to catch on a good hour. You cannot produce a real uptime percentage this way, and neither can anyone else without continuous probing, so describe what you saw honestly rather than inventing a figure.
- Confirm signing practice, not just its existence
- Check that announcements arrive signed and that the signatures actually verify against the key you obtained independently. A published key that never signs anything, or signatures nobody can check, is decoration. The point is not that a key exists; it is that you can use it to tell a genuine statement from a forged one.
- Distinguish a pattern from an anecdote on support and disputes
- One person's good experience and one person's disaster are both anecdotes. What you are looking for is enough separate, non-coordinated accounts of how support responds and how disputes resolve that a shape emerges which a single loud voice cannot have authored. If you cannot find that, the honest conclusion is that you do not know, not that it is fine.
- Weight conduct under pressure above conduct during calm
- Every project behaves well when nothing is wrong. The evidence that predicts survival is how it behaved through a hard stretch — an outage, a wave of new users, an incident hitting a peer. If the project has not been through one while you were watching, say so, and treat your confidence accordingly.
Run honestly, that checklist tends to end in one of two places: a defensible personal judgement backed by evidence you could describe to someone else, or the realistic admission that you have not gathered enough to have one. Both are acceptable outcomes. The failure is the third path — concluding "probably fine" from an accumulation of impressions that never became evidence — which is exactly the path this journal prints Unrated to avoid.
Frequently asked questions
So which one should I use?
This page cannot answer that, and any page that answers it about an unassessed project is answering from something other than evidence. What it can tell you is that Torzon's shortcomings are documented and therefore manageable, whereas choosing Mars means accepting an unmapped risk profile. Some readers will reasonably prefer a known set of flaws; that preference is a legitimate conclusion to draw from this page, and it is not the same as an endorsement.
Does Unrated mean Mars is new or obscure?
Not necessarily. The label tracks our coverage rather than the project's age or prominence. A long-established, widely discussed project can sit Unrated here simply because nobody at this desk has completed the verification work, and a project's visibility elsewhere is not a substitute for that work.
Why not use community consensus in place of your own checks?
Because consensus in this ecosystem is cheap to manufacture and slow to correct. Sentiment is treated on this site as a weak signal that can support a finding, never as one that can establish it. A grade assembled from sentiment alone would inherit every distortion in its source, and would carry the same authoritative appearance as one built from observation.
Will this comparison be updated?
It is revisited whenever either side's status changes materially, following the same criteria applied everywhere on this site: a documented incident, a sustained shift in one of the seven dimensions, or enough elapsed time that the underlying evidence has gone stale. If Mars is graded later, this page gets a real second column rather than a quiet edit.
Is it fair to compare a rated market against an unrated one at all?
Fair in the sense that matters most: the page shows the reader exactly where knowledge ends rather than papering over it. A tidy grid with both columns filled from inference would look more even-handed and be considerably less true. The empty column is not a failure of the comparison; it is the comparison's most honest content, and a reader is better served by seeing the boundary of what this desk knows than by a symmetrical table assembled to look complete.
Does Torzon's Caution make it the safer default over an unrated Mars?
For a reader who is not going to do their own verification work, a documented set of flaws is genuinely easier to manage than an unmapped one, and defaulting to the assessed option is a reasonable conclusion to draw. But the framing matters: Caution is not a safety stamp and Unrated is not a warning. What you are actually choosing between is a known risk profile and an unknown one — not between safe and unsafe — and a reader who can carry the verification burden themselves is not bound by this default at all.