Sorry but can anyone confirm this? They released more models than ever last month???

How does this keep happening? Bluesky is a platform full of very loud anti-AI people, but still, it's the single platform that drowns in slop and and everybody keeps liking these images. Again, what do you want, man? How does this keep happening?

Something's odd about the launch of Mistral Medium 3.5, and it's not their fault. Artificial Analysis paid close to $1k to benchmark the model but still defaults to showing Mistral Small 4, a much worse model, on their front page. Medium 3.5 scores way better (Deepseek V3.2 level).

The new pricing of Medium 3.5 suggests Mistral is in need of money, probably because of failed training runs or low demand. They were able to provide Devstral 2 (123B dense) at a much lower cost than Medium 3.5 (128B dense), even though it should effectively be almost identically heavy on hardware.

Mistral dropped the bomb: a 128B DENSE Medium 3.5. They indeed seemed to take a Devstral-like architecture (same hidden size, attn heads/dims, among other things), maybe even literally a Devstral checkpoint, and added a vision encoder on top. But not just that: They added reasoning 1/5

Gemma 4 E2B reaches a whopping 620 tok/s prefill speed on the dated Tensor G4 GPU! This is truly impressive. Shows that simply interleaving sliding window attention with full attention is still a worthy technique. Wouldn't have guessed that, to be honest.

New open-weights, likely multimodal+dense Mistral Medium model (128B) inbound for tomorrow, hopes are high as always! It's still unclear whether the model has reasoning baked in, as you can't see that from the architecture alone, but if it has, it could become a real banger!

Deepseek V4 finally released today, if you haven't noticed yet. I've spent some time reading their technical report and benchmarks and my feelings are extremely mixed. I am full of thoughts, but I feel like it's too early to pinpoint most things for V4 yet.

Peak deutscher Humor ist so fucking cursed dass er für jeden, der die Sprache im Nachhinein lernt, der reinste Fiebertraum sein muss, ich liebs

I am, by no means, a Trump supporter. I was shattered when I saw Kamala lose to this guy. But we shouldn't give up clean research. That said, is there any credible source for the claim on the image? I haven't been able to find anything. Genuinely curious.

This is extremely ironic. Like, comically ironic.

I usually don't post these "hype"-ish posts, but I need to say it: Qwen 3.6 35B A3B is genuinely impressive. I'm able to run it quantized to Q4_K_M @ 50+ tok/s on my RTX 4070 Ti (12 GB VRAM) with FULL (!) 256k context length. It's smart and actually useful. Local models are evolving really quickly.

Reasoning LLMs are the biggest compromise in machine learning - converting all those rich, contextual embeddings into discrete token IDs and then feeding them back. Awful. I'm confident "they" will come up with something more efficient soon. Please.

I don't like the chains of thought of the newer open-weight LLMs on the market. They just don't try to be efficient anymore. I know, it's supposed to be more structured and stuff, but I think leaving all the distillation artifacts from bigger models like the one marked in the image is unacceptable.

I came up with a tremendous, nearly undetectable method to cheat in exams; instead of trying to smuggle in an AI assistant akin to ChatGPT to help you out, simply *remember* (!) the weights of an open, frontier LLM like, say, GLM 5.1.

I welcome the two-week ceasefire the US and Iran agreed last night. It brings much-needed de-escalation. I thank Pakistan for its mediation. Now it is crucial that negotiations for an enduring solution to this conflict continue. We will continue coordinating with our partners to this end.

This is such an astronomically great thing! 🥹

Why in the name of fuck is Gemini 3.1 Flash Lite priced at $1.50/M output tokens? lil bro it's not that good 🥀🥀

I'm really, really looking forward to Deepseek V4! Let's just hope it releases soon, because the competition is evolving a lot right now...

We're happy to announce a long-term partnership with Motorola. We're collaborating on future devices meeting our privacy and security standards with official GrapheneOS support. motorolanews.com/motorola-thr...

Gemini 3 Flash (Fast mode) is literally just a reasoning model that pretends like it's not and any comparison between instruct models is inherently unfair. Even minimal reasoning is still reasoning and you can clearly feel the difference in the response quality my opinion.

Graphene Is All You Need.

Every VLM Implementation except for Qwen's and Gemini's feels botched.

I'm slightly disappointed in Mistral Large 3 being solely based on a recycled Deepseek architecture with minor changes. Mistral has a lot of potential, and while trying out custom architectures is risky for smaller ML startups, it's the only way to remain independent in the long term.

I can't help but think that Gemini 3 Flash has been even more benchmaxxed than other models... Besides that, they waited way too with publishing Gemini 3. The models are only barely SOTA a few weeks after their initial release. What was the point of all that?!

Brace, Mistral might be dropping bombshell LLMs very soon. 3B, 8B, and one proprietary.

I know the Artificial Analysis Index can be inaccurate as hell But boy, is that a hell of a beautiful sight.

YouTube is rage baiting everyone, yet again: They finally introduced a selector on MOBILE WEB for the "Audio track" which allows you to get rid of the terrible auto-translation... BUT you can only use this for Shorts. Not for full videos. They don't have that option there. How the fuck.

No app is as good at keeping you logged out as Bluesky. My mailbox is bursting more and more with dozens of log in mails and I'm tired of pretending it's not.

Okay, so Android is "de jure open source" but de facto closed source right now?? Great job, Google 😮‍💨

Yesterday, I talked with a more or less influential Shiftphone employee on the IFA about their phone and future support for @grapheneos.org. They said the question isn't new - and that they'd happily work on supporting GrapheneOS. 1/5

Had my Pixel 7a get fixed yesterday, but now I'm totally out of the loop! Did Android 16 QPR1 finally get released?? 😳 Can I expect a One UI clone in a few days?

My smartphone broke (Pixel 7a, fell on concrete) and there's nothing better than seeing GrapheneOS Attestation work more flawless than ever! Give it a shot at attestation.app!

> GPT-4 I thought you said GPT-5 was so great?

Wie bitte????? "Unbehagen"? No shit sherlock #NATO #Finland

He can't keep getting away with it.

HEAR ME OUT GUYS AND GALS THEY DID FUCKING COOK. lmarena.ai/leaderboard/... EU may be back in business

Nothing hits harder than setting the GrapheneOS auto reboot timer to 2 hours while the most sensitive thing on your phone are photos of your dog stealing a Burrito

How I feel telling my GrapheneOS duress PIN to my friend who just wanted to look something up on my phone (I'll have to reinstall everything)

Ente released PrivacyPack, a new website to collect your favorite services for increased online privacy. Make your own pack here! privacypack.org

Flux is COOKING man! (deliberately posting this before nano-banana releases) Crazy! GO, EU, GO 🔥

Met a friendly dude in IT who asked me if my phone on the desk was a Google Pixel. "Yes", I said, "I put GrapheneOS on it!" "Cool!", he pulled out his phone. "Mine is called a Fairphone. It's great. I put my custom ROM on it, but you can also install GrapheneOS on it!" On a Fairphone? Mm-hmm.

Their "smaller" (20b a3.6b) model runs with less than 15 tok/s on my 4070 Ti. Yeah I don't think it'll "run" on a phone.

Actually, this is a terrible take.

Mistral ... how's it going? Have you been cooking? It's been 2 months and the AI market has rarely been more competitive, so PLEASE act right NOW

Calling GrapheneOS an "Android build" is the understatement of the year.

I just found out something mind-blowing about Voxtral, the new European voice processing LLM by Mistral AI: Not only does it transcribe your voice with impressive accuracy, it also CORRECTS GRAMMATICAL ERRORS WHILE YOU SPEAK!! How fricking cool is that??

> We're working with a 💫 major Android OEM 💫 These cliffhangers make me go crazy, just enough tension to keep us tuning in for the next post, hoping that the ominous OEM is revealed, but NOO 😭