The destruction of the Kakhovka Dam flooded over 500 square kilometres and carried a significant death toll, caused major damage to the surrounding area and impacted over a million people.
Besides maybe solar, there is always the potential for harm during war times. Heck, even with solar, cutting off electricity supply to a major urban area can tragically carry a death toll that goes beyond even the few severe power plant disasters we as a species have ever encountered.
Fission is not an outlier here and designing any power plant for resilience is a necessary step, whether in war or peacetimes, whether in a tectonically stable or seismically active area.
Not the battery backup system. And batteries are incredibly explosive and will explode sympathetically. Its much like hitting a refinery in that regard. NPPs have been hit during this war with bombs, they survived just fine because they are built in much the same way as a hardened military structure. They can withstand a fully loaded 747 hitting them directly at the speed of sound. That's more boom than all but the largest ordinance.
A nuclear strike on a larger fission reactor could create an awful result unlike any other power plant. There were some studies during the cold war (something about the effects in West Germanyin the early 1980s, IIRC).
As opposed to a nuclear strike directly on a populated civilian area with no nuclear power plant nearby which also creates an awful result?
The main difference in nuking a fission plant over any other area would be creating longer-term fallout.
If the attacker is at all rational, unlikely to be desired because it makes the area hard to inaccessible for troops on either side.
If the attacker is horrificaly irrational and just wants to cause pure, long term, irreparable harm, unfortunately salted nukes would be a feasible option, independent of having a fission plant anywhere near the drop zone...
Nuclear disarmament, international cooperation and reliable defence alliances can (hopefully) be used to protect against this, along with resilient fission designs within reason. Not having a fission plant by itself unfortunately and targically wouldn't prevent an utterly irrational attacker willing to cause as much long term damage as possible.
Very aware of that harsh fact, living in a country that has a section in our constitution making nuclear power illegal (yes, very stupid) and is not part of any defensive alliance (while expecting the countries around us to defend us anyways), if someone wanted to "salt the earth" around Salzburg, Vienna or Graz, they could do it, even if we have no plants.
Scale of fallout would be huge. People were kind of shocked by the analysis (trying to remember what came out then... Fetter, early 1980s, Scientific American?).
They slashed pricing, it's now nearly half vs the preview (Input: $0.7/M Tokens, Output: $2.1/M Tokens vs Input: $1.36/M Tokens, Output: $4.18/M Tokens).
At the original pricing in my evals, Mistral Large 4 Preview was not competitive. With this price point and some improvements on top, this full release warrants a major reassessment.
They didn't. Original price was maybe higher but they slashed it during preview and in fact it was lower by a bit ($0.68/M tokens) and $2.09 for output.
I've been using it for my personal project and it's still too expensive and very slow when thinking. I think we need one more generation from mistral and with maybe 50% of this already "slashed" price for this to be competitive (I use Opus 5.5 at work and it's still much better than this).
I just checked, seems you are right. They had roughly a week with the lower pricing, then increased it, now decrease it again. Very odd. My testing was with the higher pricing and looking at AA, theirs seems to have been too.
Not sure it’s particularly odd. Several models of late have launched with an introductory promotional pricing, only to later lock in the promotional pricing.
Too many models now to remember specifics - I believe deepseek v4 did this and one of the gpt models (5? 5.1? 6?) did this if I recall correctly. I’m also fairly confident Google has had a few promotional periods on launch of the model, I seem to remember them starting the trend.
Token providers in aggregate work a bit like airlines so the end pricing should also work a bit like airlines; expect budget tokens, fast boarding tokens, window seat tokens, last minute tokens, miles for tokens, … and naturally prices fluctuating by the hour.
Yeah I tried mistral 4 when it came out and it honestly felt like going back in time 6 months, suddenly my agents were making all kinds of mistakes that I had kind of forgotten agents regularly used to make
> They slashed pricing, it's now nearly half vs the preview (Input: $0.7/M Tokens, Output: $2.1/M Tokens vs Input: $1.36/M Tokens, Output: $4.18/M Tokens).
Those in the EU wanting EU-only inference should be aware Mistral put a 10% tax ("upcharge") on Regional Inference[1].
Given the only practical alternative to a jurisdiction with strong data protection laws is a home rig, as the saying goes, still cheap at twice the price.
Irregardless of wether or not Mistral Large 4 is competitive or not, I do not trust the AA numbers anymore, for various reasons. Them changing their numbers around so much recently brings the entire validity of this in question, not to mention if such a combined number is even important at all. The combined "intelligence" score of AA includes areas I do not care about or need at all.
Agree. They had just fiddled with the formula a few weeks prior to this, too. There seems to be an effort to get a certain result from the model rankings.
Exactly, these general scores are pointless. Let's say the model is amazing on A but crap on B and C. It gets low score, but you still would benefit greatly if you need it for A.
> this full release warrants a major reassessment.
I really don't think it does, the benchmark numbers are dreadful and the cost is higher than Sol. If you have to use an EU model knock yourself out but there are a ton of models that are a lot smarter for a lot less money.
So if scores are good, they must be benchmaxxed, and if they are bad, they must not be benchmaxxed. So basically, the narrative is whatever you want it, no matter what the data says.
No, if a model scores good on benchmarks it could be that they have benchmaxxed, if a modern model scores bad on benchmarks it could be a really bad model but it almost certainly isn't benchmaxxed. It's relatively easy to make a model do well on particular benchmarks, companies that don't care about that stand a better chance at making a model that does well in general because tweaking a model for benchmarks usually is at the expense of something else.
I run my own benchmarks (this model is - unfortunately too large so that won't be happening here) on data that no outsider has ever seen and this gives sometimes results that are not in line with the order that you would expect if you took the various benchmark sites into account. It is of course entirely possible that our benchmark is not representative but it seems to work quite well as a predictor for which models have value for us and which do not. Recently we looked a model that has been pushed with a significant amount of hype and for the life of me we could not get it to perform even close to the published benchmarks, whereas the models we used for controls were definitely within the margin of error. It's not always equally cut-and-dry.
I don't quite believe the benchmarks, especially supplied by the company.
Qwen3.8-Flash-Next consistently benchmarks better than both GLM5.3-Flash and the previous DeepSeek v4.1 Flash 0731. The DS 0731 was so atrocious it was getting beaten by small Nemotron in my coding tests (subset of tasks from open source coding benchmarks).
Now, new DeepSeek V4.1 Flash 1017 really gets the numbers they promised for the first one, but it's too big for me to run the 4 bit quant locally at interactive speeds.
Given the size of the model, that's very aggressive pricing.
I wonder if it is serving at a loss to get training feedback (maybe using the margins they get from GLM 5.3 serving which is a slightly smaller model but double the price).
based on Anthropic's S1 no one is serving models at a loss, but cutting your price to keep usage high makes a lot of sense when you have your inference booked for long periods.
That has never been made available even to limited customers as far as reporting goes. Currently available ultrafast on GPT-6 Astra and GPT-6.1 Sol are all using Nvidia for good reason [0].
The fact that Cerebras, which has a willing and major customer like OpenAI begging at their heels to deploy on their hardware, now posts about a 2.8T open weight model says a lot.
For some reason though on openrouter the tps of 6.1 sol ultrafast is usually lower than opus 5.5 fast. Is there readings off as surely cerebras hardware is faster than low user density nvidia still?
GPT-6.1 Sol Ultrafast also uses Nvidia hardware, not Cerebras, just far better provisioned vs how they usually provide inference allowing for higher throughput.
"Much better provisioned" might be overselling it. Mainly they are just serving very few customers per Nvidia GPU, which is an expensive solution to the problem.
Can anyone explain the two month window between reporting and disclosure? I do understand Responsible Disclosure as a concept, but looking at the timeline and severity, I struggle to understand the timing, especially the month that passed before an acknowledgment.
Edit: Apparently this had been reported more than half a decade ago [0]. I am getting more confused the more I look into this, hopeful for a well written, factual post-mortem from someone who hasn't actively avoided PHP for the last decade.
> How is a bug that was patched six years ago causing RCE today?
From what I read last night: that patch fixed a specific case (an over-large index that became a negative number) that could trigger the problem, but did not fix the core problem itself. The new exploits try to trigger the problem in another manner, the new patch does address the core issue.
Can anyone explain the lack of the irresponsible disclosure in the 'Responsible' disclosure timeline? The one that happened just before the coordinated release?
Here's the missing part, as per the PHP announcement
> On October 21st at 07:41 UTC, more than six hours before the coordinated release, Chainguard unilaterally broke the embargo and published built images containing the fix. Like other distributors, Chainguard had received the patch in advance so its users could update the moment the vulnerabilities were disclosed. Any attacker could now study the fix in binary form while every other PHP user was still waiting for the release. This disregarded weeks of work by many parties toward a coordinated disclosure.
People should stop reading AI slop on X and other social media. The exploit is advanced and took some effort to fix. If there really was an open RCE in PHP for anyone to use the impact would be massive.
We don't know yet. Not sure how that changes things either way. Impact goes far beyond getting exploited and the leak wasn't ideal in that regard either.
State hackers and others prepare how to use exploits long before they happen. If there was an opened known RCE in PHP they would have caused havoc immediately. Clearly that did not happen.
Many exploits are used by security services, so this gives time to wrap any investigations or surveillance or find another method of accessing target servers.
> The AMD hardware is x86_64, which means lower battery life on average for normal users [...]
A Ryzen AI Max+ 395 powered tablet with 70Wh [0] gets the same battery life given the same task as a Nvidia N1X powered laptop with 99.9 Wh [1].
Power Consumption in their "Cyberpunk / Stress Test" is also lower at Ø109.3 (108.8-110.8) Watt for the x86_64 vs Ø134.5 (132.1-137.5) Watt for the arm64 system using the same settings, resolution and display brightness. They also get the same performance when normalising for resolutions and settings in Cyberpunk, so this is not an instance of one providing significantly more performance.
I suspect this is going to be a Cybertruck/Boring Company situation, where we may get something ostensibly called the same and may even do something one could consider similar when squinting (maybe a bit higher than U9 jumps in scope) but with the specs and implementation differing so greatly that it shouldn't be considered the same. Exoskeleton vs body panels, 200kph+ on electric sleds vs Teslas (driven by human operators; tells you how much they trust their FSD tech) in a tiny tunnel, FSD on HW3 vs an upgrade maybe someday, etc.
And inevitably two groups will clash, one pointing out how much this deviated from the original, often decade long promises vs another pointing at the identical naming, how it's still a product they like and of course the all time winner "I don't see you do X"...
For what it's worth, I'd file this as a RCS demo for SpaceXGrokSI (or whatever they call themselves this week) above all else and if I were a betting man, might be a prelude to fusing them into SpaceXSuperXTeslaXIntelligenceX...
I used to feel the same way, namely that the tunnel driving would be the signal that unsupervised driving was ready to go. But now that I see their geofenced robotaxis in operation I sort of get it now. Those things do well at driving autonomously, but there are edge cases where the car's safety response is to essentially shut down and wait for human intervention rather than risk uncertainty. That's not something you want happening when you have a single lane of traffic with no passing, leading to a situation where every car in the chain is now forced to deal with the situation in unison. Humans can figure that out pretty quickly, but training the cars to do it will take time. I do expect autonomous tunnel rides any day now, but I can see why they weren't the first use case for autonomous vehicles.
Legal argument is already a strong one, but I feel that the training data in there must be invaluable. Such a mass of shipped features, regressions, refactors, complete rewrites and pivots, all with full internal history, is a goldmine. Whether archiving and switching branch might be in order, that's a fair discussion though.
Besides maybe solar, there is always the potential for harm during war times. Heck, even with solar, cutting off electricity supply to a major urban area can tragically carry a death toll that goes beyond even the few severe power plant disasters we as a species have ever encountered.
Fission is not an outlier here and designing any power plant for resilience is a necessary step, whether in war or peacetimes, whether in a tectonically stable or seismically active area.
reply