How Many Died in the Twelve-Day War?

The most cited fact about the June 2025 war is the one number nobody can state cleanly. Ask how many people died in the Twelve-Day War between Israel and Iran and you will get, depending on who you ask and when you asked them, 28, or 610, or 935, or 1,062, or 1,190. All of these are real, sourced figures. None of them is wrong in the way a fabricated number is wrong. They are answers to slightly different questions asked of slightly different evidence, and the distance between them is not noise to be averaged away. It is the most honest thing the casualty record holds.

This briefing is an attempt to say what is actually known about the human cost of the war, and, just as carefully, what is not. The exercise matters because a death toll is the rare fact that is simultaneously the most emotionally weighted claim a war produces and the most methodologically fragile. Everyone reaches for a single number. The single number is almost always a political artifact: a count performed by an interested party, under conditions that shaped what could be seen, and released at a moment chosen for effect. On a fact-based record like Blackwing, the toll is not a headline to settle. It is a range to hold, with its uncertainty intact, because the shape of the uncertainty is where the argument actually lives.

The short version is this. Israel's toll is close to settled: 28 killed. Iran's is not, and it is instructive that it is not. The official Iranian figure climbed for six weeks after the guns went quiet, from a mid-war 224 to a final 1,062. Two independent human-rights monitors, working from field documentation rather than state announcements, arrived at 1,190 and 1,082. The three most serious counts, in other words, converge on a total near 1,100 dead in Iran, and then diverge sharply on the single figure that carries the most legal and moral weight: how many of those dead were civilians. That number ranges from 182 to 436 across the three counts, a spread of more than two to one. Everything hard about this war's human cost is compressed into that gap.

1. Why "the death toll" is a category error

Start with the structural mistake almost every reader makes, including careful ones. We treat a death toll as an observed quantity, like a temperature or a stock price: a thing that exists in the world and merely needs to be read off. It is not. A death toll is a count that someone performs, and every count is the product of three things that are easy to forget once the number is printed. First, who is doing the counting, and what they want the number to say. Second, what the counter can actually see, which in a war is never everything and in a closed society is often very little. Third, how each death is classified, because the category a body is assigned to (civilian, combatant, unconfirmed) changes the political meaning of the whole.

This is not a nihilist point. It does not mean all numbers are equally good or that the truth is unknowable. It means the opposite: that to know anything real, you have to stop looking for the number and start looking at the counts, plural, and at the machinery that produced each one. A single figure with no provenance is worth less than a range with a documented shape, because the range tells you where the counters agree (and are therefore probably right) and where they diverge (and are therefore telling you something about the limits of the evidence, or about their own incentives).

There is a reason the toll, of all the facts a war produces, attracts the most manipulation, and naming it sharpens everything that follows. A death toll is not just a measurement; it is a claim about who the victim is. A high civilian count indicts the attacker; a high military count vindicates them; a low toll on your own side projects competence; a high toll on the enemy's projects resolve. Every party to a war therefore has a preferred number, and the preferred numbers point in incompatible directions, which is precisely why the same event yields a spread and not a point. This is not cynicism about the possibility of truth. It is realism about the forces acting on the truth: a toll is contested in proportion to how much rides on it, and in a war fought partly to justify itself to the world, almost everything rides on it. A reader who forgets that the number is also a weapon will keep mistaking a rhetorical instrument for a neutral gauge, and will keep being surprised when the instrument reads differently in different hands.

The war made this abstract point concrete in an unusually clean way, because it produced two counting environments that could hardly be more different. On one side, an open democracy with a free press, a health ministry that is not a party to concealing its own dead, and a public that expects to be told the names. On the other, an authoritarian state that treats casualty figures as a controlled resource, that revised its own count upward for weeks, and whose final tally was contested in real time by independent monitors who put the civilian share at nearly double the official figure. The same war, counted two ways, is a natural experiment in how a number is made, and reading the two side by side is the fastest route to understanding why any war's toll should be read as a claim, not a reading.

The rest of this briefing works through the evidence in that spirit. It takes the settled side first, because the contrast sets up everything after it. Then it follows Iran's number as it moved, examines the three serious counts and the argument buried inside them, and drills into a single strike, on Evin prison, where the whole methodological problem plays out in one building with four different numbers attached to it. Along the way it marks the places where the record simply gives out: the nuclear-site strikes, where no casualty figure was ever published and the honest entry is not zero but unconfirmed. The goal is not to hand you a total. It is to hand you the boundary between what is known and what is asserted, and to be honest about where that boundary falls.

2. Israel's 28: what a settled number looks like

Twenty-eight people were killed in Israel by Iranian missiles over the twelve days - The Times of Israel. That figure is about as close to settled as a wartime toll ever gets, and it is worth understanding why it is settled, because the reasons are exactly the ingredients Iran's count lacks. The Israeli dead were named. The oldest was a 95-year-old Holocaust survivor; the youngest, a 7-year-old Ukrainian cancer patient who had come to Israel for treatment. Their deaths were recorded by hospitals, reported by a press that is free to report them, and cross-checked against a health ministry and the emergency service Magen David Adom, neither of which had any institutional reason to undercount. When a number can be resolved to a list of names, the space for dispute collapses. You cannot argue a named 7-year-old into or out of existence.

The wounded figure is fuzzier, as wounded figures always are, but only at the edges. Medical officials reported 28 killed and more than 3,000 wounded during the conflict - The Times of Israel, with fuller accountings placing hospitalizations closer to 3,200 to 3,400. The variance there is instructive precisely because it is small and boring: it reflects the ordinary difference between "treated at the scene" and "admitted," not a contested political fact. Nobody is fighting over whether Israel's wounded were combatants. The number wobbles by a few percent and settles, which is what a number does when the counting environment is honest and the incentives are neutral.

Set that against the scale of the bombardment, and the smallness of 28 becomes its own kind of finding. Iran fired roughly 550 ballistic missiles and around 1,000 drones at Israeli cities over the twelve days - FDD's Long War Journal, the largest sustained missile campaign ever directed at the country. Most were intercepted. Israeli and American air defenses stopped the large majority, at an interception rate Israel's Defense Ministry later assessed at about 86 percent - The Times of Israel. The 28 deaths are what the other fourteen percent did. That ratio, 1,550 projectiles and 28 dead, is not a story about Iranian restraint. It is a story about layered interception, hardened shelters, and a warned population, and it is the reason the human cost of this war fell so overwhelmingly on one side. Hold that asymmetry in mind, because it is the fixed point against which Iran's moving number has to be read.

Where those 28 died refutes any notion that the low toll meant Iran aimed only at soldiers. The single deadliest strike of the war hit an apartment building in Bat Yam, south of Tel Aviv, killing nine civilians and wounding more than 100 - Al Jazeera. On 14 June a missile struck a home in Tamra, a Palestinian town in the Galilee, killing four women of one family, the wife, two daughters, and sister-in-law of a local lawyer - The Times of Israel. Other missiles hit residential Ramat Gan and scored a direct hit on the Soroka hospital in Beersheba, wounding dozens. The dead were overwhelmingly civilians in their homes, which is why Human Rights Watch later concluded that Iran's missile strikes on Israeli population centers were likely war crimes - Human Rights Watch. The smallness of the Israeli number is not a story about whom Iran aimed at. It is a story about who survived being aimed at, and the difference is measured in shelters and interceptors, not in restraint.

The deeper lesson of the Israeli count is methodological, and it travels. A number is trustworthy in proportion to the openness of the environment that produced it. Free press, named victims, a counting institution with no stake in the answer, and a total small enough to resolve to individuals: those conditions make 28 durable. Strip them away, one by one, and you get Iran's number, which had none of them and behaved accordingly.

3. Iran's moving number: a toll that rose for six weeks

If Israel's 28 shows what a settled number looks like, Iran's toll shows what an unsettled one does: it moves. The official Iranian figure did not arrive as a fact and stay put. It climbed for six weeks, and the trajectory of that climb is itself evidence about the conditions underneath it. Mid-war, on 15 June, Iran's Health Ministry put the dead at 224 and stressed that the majority were civilians. By the ceasefire on 24 June, the official figure had reached 610 - Al Jazeera. The next day it was 627 - GlobalSecurity.org. Four days after that, on 28 June, Tehran raised it to 935 - The Times of Israel. The final official count, given weeks later, was 1,062 - Iran International. The number nearly quintupled from its mid-war value to its final one.

There are two innocent reasons a toll rises after a war and one less innocent one, and all three are probably at work here. The first innocent reason is recovery: the dead are pulled from rubble over days and weeks, and a strike that collapses a building does not yield its full count on the first afternoon. The second is death from wounds: the severely injured die later, and the war produced thousands of wounded. Both of these would push any honest count upward over time, and neither implies bad faith. The less innocent reason is the one specific to a closed information environment: the state controls the count, and control is exercised in stages. A government that treats casualty figures as a managed resource does not release the full number at once. It releases what is convenient when it is convenient, and the shape of that release, low and stabilizing during the fighting, then rising once the ceasefire held, is consistent with a count being administered rather than merely tallied.

We do not need to adjudicate motive to draw the practical lesson, and this is where a fact-based record has to be disciplined. The mid-war figure was not the toll; it was the toll the state was willing to confirm mid-war. Any consumer of wartime casualty data who anchors on the number available during the fighting will systematically undercount, because the number available during the fighting is, by construction, the earliest and lowest one. This is why the live map at Blackwing treats the human cost as a quantity that changes as you move through time: each official figure, the mid-war 224, the ceasefire's 610, then 935 and 1,062, appears as its own dated, sourced mark you can scrub to, so the staged release is visible as it happened. What the map deliberately does not do is stack those announcements into one running total, because each one restates the war's entire count and adding them together would tally the same dead over and over. The running counter beneath the scrubber holds to the deaths a specific dated strike can be attributed, and lets each announcement stand as the report it is. A toll is not a fact stamped on a war. It is a process that resolves over weeks, and a record that pretends otherwise is lying by omission.

It helps to be concrete about the mechanics of a staged count, because they are neither mysterious nor uniquely Iranian, only more pronounced in a closed system. When a missile levels a residential block, the dead are not all found at once: rescue crews work for days, and the confirmed count rises as bodies are recovered, a process reported transparently in an open society and selectively in a closed one. Layered on top of recovery is death from wounds, which pushes any toll upward for weeks after the last shot. What makes Iran's case distinctive is not that these mechanisms operated, they operate in every war, but that a single interested authority controlled the release of the resulting figures and had both the means and the motive to meter them. In an open system, competing hospitals and a free press force the number out roughly as fast as it accumulates. In Iran's, it emerged at the pace the state chose, which is why it stabilized reassuringly during the fighting and then climbed once the political cost of a higher figure had fallen. The trajectory is the tell: honest counts rise because bodies are found; managed counts rise because the calendar changed. Iran's did both at once, and no outside observer can fully separate the two, which is itself a finding about the limits of the record.

The revision also reframes what "official" means. By the end, Iran's own government had settled on 1,062, more than one and a half times its own ceasefire-day figure. So the gap that matters is not, as the early coverage implied, between a low Iranian claim and a high Western one. It is between two high counts that agree on the order of magnitude, Iran's official 1,062 and the independent monitors' 1,082 and 1,190, and disagree about something more specific and more damning. To see what, you have to stop looking at the totals and look at how each count sorts the dead.

4. Three counts, one argument: the civilian share

Here is the finding that a careless reading of this war misses entirely. Three serious counts of the Iranian dead agree, remarkably, on the total, and disagree, profoundly, on the split. Iran's official final figure is 1,062 killed: 786 military and 276 civilian, a tally that also recorded 102 women and 38 children among the dead, with Tehran province hardest hit at 265 - Iran International. The Washington-based monitor Human Rights Activists in Iran (HRANA) documented 1,190 killed: 436 civilians, 435 military, and 319 whose status it could not confirm, plus more than 4,000 wounded - HRANA. The Kurdish-focused rights group Hengaw counted 1,082 killed: 900 military or security personnel and 182 civilians, including 68 women and 45 children - Hengaw.

Line those up and the convergence on the total is striking. 1,062, 1,082, 1,190. Three counters, using three different methods, working from three different vantage points, land within roughly ten percent of one another on how many people died. That agreement is not an accident, and it is the strongest single thing the casualty record establishes: the war killed on the order of eleven hundred people in Iran, and any figure far below that (the mid-war 224, the ceasefire-day 610) should be understood as an early undercount, not a competing truth.

Now look at the civilian column, and the agreement detonates. Iran official: 276 civilians. Hengaw: 182 civilians. HRANA: 436 civilians. The most authoritative-sounding number, the state's own, and the two independent human-rights counts do not merely differ at the margin; HRANA's civilian figure is more than double Hengaw's and roughly 60 percent above Iran's. On the single number that determines whether this was a war fought largely between combatants or one that killed hundreds of civilians, the counters split by a factor of nearly two and a half. The total is roughly known. The split is where the truth fragments. And the split is not a rounding problem. It is the number that decides the war's moral and legal character, which is exactly why it is the number no one can pin down.

The convergence on the total and the divergence on the split are, in fact, the same phenomenon seen twice. A total is a robust quantity. It leaks out through many independent channels: funerals, hospital records, obituaries, family accounts, satellite views of fresh graves, the simple arithmetic of named strikes on known buildings. No single actor can suppress all of those channels, so three counters working independently drift toward the same magnitude. A classification is a fragile quantity. It exists only in the act of judgment, one body at a time, and that judgment depends on information (Was this person Basij? Was that engineer working on weapons?) that is far scarcer than the fact of a death and far more contestable. So the same evidence that fixes the total to within ten percent leaves the split wide open, and no amount of care closes it, because the missing information is not missing by accident.

One quantity is even less settled than the dead, and it is worth pausing on because it is usually ignored: the wounded. HRANA documented more than 4,000 injured in Iran alongside its 1,190 dead - HRANA, and other tallies ran higher, with one Iranian accounting putting the total killed or injured above 6,000 - GlobalSecurity.org. Wounded figures are always softer than death tolls, for a structural reason: a death is a discrete, countable event, while a wound is a spectrum, from a scratch treated at the scene to an injury that kills weeks later. The threshold of what counts as "wounded" is fuzzy in a way the fact of death is not, and in a closed system with overwhelmed hospitals it is fuzzier again. This matters because the wounded are where the delayed dead come from: a war's final toll is not fixed at the ceasefire but keeps rising as the severely injured succumb, which is one more reason Iran's number climbed for weeks, and one more reason every figure should be read with its date attached, never as a timeless fact.

5. Why the split is the entire argument

To understand why the civilian share is contested, you have to see that each counter is answering a slightly different question, and each has a reason to answer it the way they do. This is not a claim that anyone is lying, though in a closed information environment that possibility can never be dismissed. It is a claim about structure: the three counts diverge because the boundary between "civilian" and "military" is genuinely blurry in this war, and because each counter resolves that blur in a direction that fits their vantage and their interest.

Consider what Iran's official framing accomplishes. A toll of 786 military and 276 civilian tells a specific story: that Israel and the United States waged war on Iran's armed forces and its nuclear and missile establishment, and that the state absorbed the blow to its soldiers and scientists while sparing most of its people. For a government whose legitimacy rests on projecting strength and deflecting the charge that it exposed its own civilians, a military-heavy split is the politically survivable count. It is also, crucially, not obviously false: Israel did concentrate its opening strikes on commanders, nuclear scientists, and military infrastructure, and a genuinely large share of the early dead were in fact military. Hengaw's independent count, 900 military to 182 civilian, actually leans further in that direction than Iran's own, which is a useful check: the military-heavy shape is not merely Iranian propaganda, because a Kurdish rights group with every reason to highlight civilian suffering also found the dead to be overwhelmingly military.

Then consider HRANA's 436 civilians, nearly half its total. HRANA is not counting differently out of carelessness; it is counting differently because it classifies differently and confirms differently. Its 319 "unconfirmed" cases, a category the other counts largely lack, is the tell. HRANA is explicitly holding open a large bucket of deaths it cannot yet sort, and it is willing to place more of the identifiable dead on the civilian side of a line that Iran and Hengaw draw elsewhere. Where exactly is that line? It runs through the war's most awkward categories. A nuclear scientist killed in his apartment: civilian or combatant? A Basij militiaman, part of a paramilitary volunteer force embedded in civilian life: soldier or citizen? A technical worker at a missile facility? A prison guard? Each counter draws the line at a different place, and because the war's dead cluster heavily in exactly these ambiguous roles, small differences in where the line falls produce large differences in the civilian total.

To see why that line is so hard to draw, look at whom Israel killed in the war's opening hours, because the answer is the classification problem made flesh. Operation Rising Lion did not begin with a broad bombardment; it began with a decapitation, a synchronized set of strikes on the men who ran Iran's military and nuclear establishment. By the end of the first day Israel had killed at least 20 senior commanders and scientists - FDD. Among them were the two highest uniformed officers in the country: Major General Mohammad Bagheri, chief of staff of the armed forces and the second most powerful man in Iran after the Supreme Leader, and Major General Hossein Salami, commander in chief of the Islamic Revolutionary Guard Corps. Also killed were Gholam Ali Rashid, deputy commander of the armed forces, and Amir Ali Hajizadeh, head of the IRGC's aerospace force and the architect of Iran's missile program - TIME. These are combatants by any definition, and every count places them on the military side of the ledger.

The scientists are where the definition breaks. Israel also killed a cluster of nuclear physicists in their homes, among them Fereydoun Abbasi-Davani, a former head of the Atomic Energy Organization of Iran, and Mohammad Mehdi Tehranchi, a physicist who led a Tehran university - TIME. Were they combatants? Israel's case for striking them rests on the claim that they were essential inputs to a weapons program, a military-necessity argument, not a claim that they wore a uniform. A rights monitor applying the letter of humanitarian law might record a physicist killed in his apartment as a civilian death, because he was not a member of an armed force and was not, at the moment of his death, taking direct part in hostilities. Iran, for its part, folded the scientists into its state funeral for fallen commanders on 28 June, honoring them as martyrs of the armed struggle - Al Jazeera, which is to say Iran itself classified them as military when it was politically useful to do so. The same dead man is a civilian to a rights lawyer, a combatant to Israel's targeting cell, and a martyr-soldier to the Iranian state. His body did not change. The category did, three times, according to who was classifying and why. Multiply that ambiguity across hundreds of Basij militiamen, technical staff, and security personnel, and the two-to-one spread in the civilian count stops looking like sloppiness and starts looking like the honest signature of a genuinely undecidable question.

This is the deep reason the split cannot be resolved by better data alone, and it is worth stating plainly because it is the intellectual core of the whole subject. The civilian-combatant boundary is not a fact waiting to be measured; it is a judgment that different institutions make differently, on purpose. International humanitarian law has its own definitions, human-rights groups apply stricter ones, states apply self-serving ones, and none of them can be "checked" against a ground truth, because the ground truth is a legal category, not a physical one. When you read that "hundreds of civilians died," the honest follow-up is never "is that correct?" It is "by whose definition of civilian, and what did they do with the scientists, the militiamen, and the unconfirmed?" That question, not the total, is where a serious reader spends their attention.

The practical upshot for the record is a discipline about language. It is defensible to write that roughly 1,100 people were killed in Iran, because three counts support it. It is not defensible to write that "most were civilians" or "most were military" as a settled fact, because the three counts split almost evenly across that claim. The most one can say honestly is that the civilian dead number in the hundreds, somewhere between about 180 and 440, and that the true figure depends on a definitional choice the sources do not share. A record that states the range and names the disagreement is telling the truth. A record that picks one split and prints it as fact is laundering a political judgment into a number, which is precisely the failure a fact-based platform exists to prevent.

6. Evin prison: the whole problem in a single strike

If the three national counts show the counting problem at scale, a single event shows it in miniature, and the Israeli strike on Evin prison on 23 June is the sharpest example the war produced. Evin is the notorious detention complex in northern Tehran long associated with political prisoners and dual nationals. Israel hit it late in the war, destroying part of its administrative building. And then the same thing happened to this one strike that happened to the whole war: it acquired not one death toll but several, and the gaps between them are a fractal of the gaps between the national counts.

Iran's judiciary spokesman said the strike killed at least 71 people, including staff, soldiers, prisoners, and family members who had come to visit - France 24. Some official accounts put the figure at 79. HRANA, counting in its own way, identified the dead more specifically: at least 35 staff members and 2 inmates, along with a person walking near the prison and a woman who had come to see a judge about her imprisoned husband. Hengaw, applying its stricter field-verification standard, confirmed only 44. One strike, one building, one afternoon, and the count runs from 44 to 79 depending entirely on who is doing the confirming and what threshold of evidence they require before they will write a death down.

Notice that the disagreement here cannot be blamed on the fog of a nationwide war. This is a single, static, well-known location, and the counters still split by almost a factor of two. That tells you the divergence is not about the difficulty of aggregating a country's worth of strikes. It is intrinsic to the act of counting the dead under these conditions: verification standards differ, access differs, and the incentive to confirm or withhold differs, even for one building. If the count for Evin cannot be nailed down to a single number, the national count certainly cannot, and anyone who quotes the national total to the nearest person is claiming a precision the Evin case proves does not exist.

The human texture of the Evin dead is worth holding onto, because it is exactly what a bare number erases. The people killed were not a category; they were staff, soldiers, inmates, and visitors. Among them, HRANA identified a woman who had come to the prison to meet a judge about her imprisoned husband's case, and a passerby walking near the walls when the missile hit - Wikipedia. A strike on a prison does not kill an abstraction called "the regime." It kills the clerk, the guard, the prisoner, and the wife visiting the prisoner, indiscriminately, because a bomb does not read the roster. This is why the classification problem at Evin is so acute: the building held, at the moment it was struck, a population that was simultaneously part of the state's apparatus of repression and among its victims, and no single label fits a toll drawn from both. The number, 71 or 44 or 79, flattens all of that into a scalar. The record's job is to keep the scalar honest while remembering that it stands for people the state itself categorized in opposite ways, employing some and imprisoning others, inside one wall that one missile did not distinguish.

The Evin strike also introduces the dimension that pure counting leaves out: not just how many, but whether they should have been killed at all. Among the confirmed dead was Ali Ghanaatkar, a top prosecutor whose pursuit of dissidents, including the Nobel laureate Narges Mohammadi, had drawn years of human-rights condemnation. His death complicates any simple civilian-versus-military frame: a prosecutor is not a soldier, but neither is he an uninvolved bystander. And the strike as a whole drew a call from Amnesty International for it to be investigated as a war crime, on the grounds that a prison full of detainees, staff, and visiting families is not a lawful military target - Amnesty International. The number, in other words, is entangled with the legality, and the legality is entangled with the classification, and the classification is the thing the counters cannot agree on. Evin is the war's casualty problem compressed into one address.

7. The strikes with no count, and the one real zero

There is a class of deaths this war produced that has no number at all, and how a record handles that absence is a test of its honesty. The centerpiece is the American strike on the nuclear sites, Fordow, Natanz, and Isfahan, on 21 and 22 June. Iran said the facilities had been evacuated in advance. No source has ever published a casualty figure for those specific strikes. The temptation is to record that as zero, and the temptation must be resisted, because a facility that Iran says was evacuated is not the same as a facility confirmed to be empty. The honest entry is not zero. It is unconfirmed, and the distinction is the entire difference between a measured fact and a convenient assumption. Reporting "US strikes: 0 killed" next to "Israel's campaign: over 1,000 killed" would be false precision: it would treat an absence of evidence as evidence of absence, and it would flatten a genuine unknown into a tidy zero for the sake of a clean-looking table.

The evacuation claim deserves scrutiny precisely because it is so convenient to everyone, which is a warning sign, not a reassurance. Iran had every reason to say the sites were emptied in advance: an evacuated facility means a strike that killed no one, which denies the attacker a clean propaganda win and spares the state the embarrassment of having failed to protect its most sensitive workers. The United States and Israel, for their part, had little interest in contesting it, because a strike that killed a hall full of scientists would invite exactly the war-crime scrutiny that a strike on empty centrifuges avoids. So the two sides that agree on almost nothing both found the "no casualties" framing useful, and a fact that every party finds convenient is a fact no party is auditing. That does not make it false: Iran may well have moved personnel out of Fordow and Natanz ahead of a strike it had seen coming for days. But "may well have" is not "did," and a record that writes zero where the evidence supports only unconfirmed has quietly upgraded a convenient assumption into a measured fact. Refusing that upgrade is the whole discipline.

This is more than a bookkeeping nicety, because the nuclear-site strikes are the ones whose consequences the world argued about most fiercely, and the argument was almost entirely about damage, not death. We examined that question at length in our briefing on what really happened to Fordow, where four intelligence agencies gave four different answers about how far the strikes set Iran's program back. The casualty question at those sites is even more opaque than the damage question, because the damage at least left seismic signatures and satellite craters, while the human cost left nothing that could be measured from outside. The record's job at Fordow is the same as its job everywhere else, only starker: mark the boundary of the evidence, and refuse to invent a number on the far side of it. Unconfirmed is not a failure of the record. It is the record being accurate about its own limits.

Against all of this sits the war's one genuinely clean zero, and it is worth stating because it shows what a real zero looks like as opposed to a manufactured one. When Iran fired its retaliatory salvo at the Al Udeid air base in Qatar on 23 June, a barrage the US had been warned about and defended with the largest Patriot engagement in American history, no US personnel were killed. That zero is real. It is confirmed by an open military that reports its own dead, it concerns a discrete and observable event, and it has never been seriously contested. The contrast with the nuclear sites is the whole point. Al Udeid is a measured zero. Fordow is an unmeasured unknown. They look identical if you only read the totals, and they are opposites if you read the provenance, which is exactly why a record has to carry the provenance and not just the number.

8. Why war tolls diverge everywhere, and more so here

Step back from the specifics and the divergence in Iran's count stops looking like an Iranian peculiarity and starts looking like an instance of a general law. Counting the war dead is a contested political act in every conflict, open or closed. It was contested in Iraq, where estimates ranged across an order of magnitude depending on method. It is contested in Gaza, where the argument over the ministry's figures is really an argument over the combatant-civilian split, the same argument we have just watched play out over Iran. The reason is structural and it is worth stating from first principles: a death toll sits at the intersection of a physical fact (a person is dead) and a set of judgments (who they were, who killed them, whether it was lawful), and only the physical fact is stable. Everything downstream of it, the classification, the attribution, the legality, is a claim, and claims are where interests operate.

What the Iranian case adds is a closed information environment layered on top of that universal problem, and the layering compounds the uncertainty in a specific, direction-having way. In an open society, the multiple channels through which a death leaks out, free press, independent hospitals, named obituaries, are numerous and mutually correcting, which is why Israel's 28 is solid. In a closed society, those channels are fewer and controllable, so the count depends more heavily on the state's own releases (which are staged and interested) and on the courage and access of independent monitors (which are limited and dangerous). The predictable result is what we observed: a total that the state could revise upward for weeks because no independent channel could force the full number out sooner, and a civilian split that independent monitors could contest but not resolve because the classifying information was locked up with the classifier.

There is a temptation, faced with this, to throw up one's hands and treat all the numbers as equally unknowable propaganda. That is the wrong lesson, and it is worth being precise about why, because the right lesson is the whole discipline of a record like this one. The numbers are not equally uncertain. The total is well-constrained (three independent counts within ten percent), the Israeli toll is nearly exact (named victims, open counting), the US personnel toll is a confirmed zero, and the civilian split and the nuclear-site toll are genuinely open. A serious record does not collapse those different grades of certainty into a single shrug. It grades them, states each at the confidence its evidence warrants, and reserves the word "unconfirmed" for the places that have actually earned it. The failure mode is not too much skepticism; it is undifferentiated skepticism, which is just as lazy as undifferentiated credulity and lets the genuinely knowable dissolve into the genuinely unknown.

The legal dimension deserves its own note, because it shows that the counting problem and the accountability problem are the same problem in two hats. Both sides' strikes drew credible war-crime allegations, from the same kind of institutions. Amnesty International called for Israel's strike on Evin prison to be investigated as a war crime - Amnesty International; Human Rights Watch concluded that Iran's missile strikes on Israeli population centers were likely war crimes - Human Rights Watch. Both judgments turn entirely on classification: whether a target was military or civilian, whether a death was of a combatant or a protected person. The war-crime question and the civilian-count question are not two subjects. They are the same act of judgment applied to the same ambiguous dead, once for a tally and once for a tribunal. This is why the split matters more than the total, and why interested parties fight over it so hard. The total is a number. The split is a verdict, and everyone counting knows it.

None of this is unique to Iran, and it helps to say so, because the reflex is to treat a contested Iranian number as evidence of Iranian bad faith specifically, when it is mostly evidence of a general law of wartime counting. The Iraq war produced casualty estimates that differed by an order of magnitude, from the roughly hundred thousand documented by passive tallies of reported deaths to the several hundred thousand implied by household mortality surveys, and the gap was a methodological argument, not a factual one: two instruments answering two different questions. The long dispute over Gaza's toll is, at its core, the same civilian-versus-combatant argument we have just watched consume the Iranian count. The pattern repeats because its cause is structural: a death toll is a physical fact wrapped in a stack of judgments, and the judgments are where politics enters. What varies from war to war is only how open the counting environment is, which sets how far apart the honest counts can drift. Iran's environment was closed, so its counts drifted far. That is the whole explanation, and it needs no special theory of Iranian mendacity to account for a spread that open societies produce, in milder form, for every war they fight.

This is the same method that runs through the rest of the platform. The intelligence layer applies it to the war's alternative-data signals, cross-referencing what seismographs, internet-outage monitors, and market tapes recorded against what the wire copy claimed, and we walked through seven of those signals in the war read through its data. The casualty question is the same exercise aimed at the hardest possible target: a number that everyone wants, that no one can fully see, and that means something different depending on who counts it. The answer is not a figure. It is a calibrated range with the uncertainty attached, and the attachment is not a hedge. It is the finding.

9. What the record can establish, and where it stops

Pull the threads together and the honest account of the war's human cost is not a number but a structured set of claims graded by confidence, which is the only form an honest account can take. It is worth laying them out explicitly, because the value of the exercise is in the grading, not in any single figure.

What is well established. Israel lost 28 people, a figure resolved to named individuals and effectively settled - The Times of Israel. Iran lost on the order of 1,100 people, a total supported by three independent counts (1,062, 1,082, and 1,190) that agree to within roughly ten percent - HRANA, Hengaw, Iran International. No US personnel were killed, a confirmed and uncontested zero. And the human cost was profoundly asymmetric, roughly forty Iranian dead for every Israeli one, a ratio that survived Iran firing more than 1,500 missiles and drones at Israeli cities.

What is genuinely contested. The civilian share of the Iranian dead, which ranges from 182 to 436 across the three serious counts and turns on a definitional judgment the sources do not share. The Evin prison toll, which runs from a field-verified 44 to an official 79 for a single strike. And the full independent-verified count, which HRANA itself brackets with 319 unconfirmed cases it was unwilling to sort. These are not gaps that more time will necessarily close, because the missing information (who exactly the dead were, and how to classify the ambiguous) is held by parties with reasons not to release it.

What the record cannot establish at all. The casualty count of the US strikes on Fordow, Natanz, and Isfahan, for which no figure was ever published and the honest entry is unconfirmed, not zero. This is the one place where the temptation to invent a clean number is strongest and must be most firmly refused, because a facility Iran called evacuated is not a facility confirmed empty, and the difference between those two statements is the difference between a record and a press release.

That is the answer, and its shape is the point. How many died in the Twelve-Day War? Twenty-eight in Israel, about eleven hundred in Iran, with a civilian toll somewhere between one hundred eighty and four hundred forty, an unknown number at the nuclear sites, and a real zero among American forces. A reader who wanted a single figure will find that unsatisfying, and their dissatisfaction is worth sitting with, because it is the exact feeling that pushes people toward the false comfort of a single number in the first place. The single number would be cleaner. It would also be a lie, or at least a claim dressed as a fact, with the disagreement quietly deleted. The range is uglier and truer, and holding it, with the confidence of each part marked and the boundary of the evidence drawn, is not a failure to answer the question. It is the answer. You can watch the toll resolve over time, event by sourced event, on the live map and through the full timeline, and read how each figure is verified in our methodology. The number moves because the truth did, and a record that refuses to move with it is not being precise. It is being false.

This briefing reflects the casualty record of the June 2025 Twelve-Day War (13 to 24 June 2025) as documented through late 2025, and does not concern any later conflict. Casualty figures for closed information environments are revised over time; every figure here is sourced to its origin, and the ranges are stated deliberately. Where the evidence does not support a number, the record says so.