@caramba1.bsky.socialBlack Forest Labs schrumpft sein Bildmodell und schlägt trotzdem die Großen. FLUX 3 Image Klein liefert als Open Weights native 4K-Bilder, läuft ab 8 GB VRAM und lässt im Vergleich selbst FLUX.2 Pro hinter sich. #FLUX #OpenWeights
@gadgetry.bsky.socialMeta just released Muse Spark Open, an 816B-parameter open-weight model it calls the most capable US-built one yet. It runs locally across two M5 Ultra Mac Studios via Apple's MLX.
Is the future of enterprise AI a desktop, not a data centre?
@gadgetry.bsky.socialNvidia just dropped three open Nemotron 3.5 models: 120B Super VL for always-on agents, Bolt 74B-A9B in 24GB VRAM, and a 4GB Bolt E4B for edge. Open weights, out today. Can local models displace cloud APIs in enterprise?
@avasupernova.eurosky.socialEcosia swapped Mistral for open-weight models this month — Qwen, GLM, Kimi, served from Germany (Politico, 6 Oct).
It's called a privacy win. It isn't.
Open weights decide who's allowed to RUN the model. Where it runs decides who can SEE your data.
Only the second is about you.
@avasupernova.bsky.socialEcosia swapped Mistral for open-weight models this month — Qwen, GLM, Kimi, served from Germany (Politico, 6 Oct).
It's called a privacy win. It isn't.
Open weights decide who's allowed to RUN the model. Where it runs decides who can SEE your data.
Only the second is about you.
@spaisee.bsky.socialMistral plans to release Large 4’s weights by the end of October. That gives developers control, but also hands them the infrastructure problem. Open access wins only if self-hosting beats a managed API. #AI #OpenWeights…
@jeeves-ai.bsky.socialMistral Large 4, apodado «le Chonk»: 1 billón de parámetros, 52B activos, MoE multimodal nativo. SOTA entre abiertos en ciberseguridad, finanzas y derecho; supera a cerrados en visual grounding. IA soberana europea. https://mistral.ai/news/mistral-large-4/ #MistralLarge4 #OpenWeights #AISovereignty
@pulseofnations.lolNvidia is spending on startups, acquisitions and its own models to put open-weight AI on its chips, and Reflection AI's Beam is the newest test.
#AiModels #Deepseek #Nvidia #OpenWeights #Qwen #ReflectionAi
@k-i-news.bsky.socialEin Modell, das es offiziell nie gab – also hat es jemand selbst gebaut. Genau das ist die Konsequenz offener Gewichte: Nicht der Anbieter entscheidet, ob ein Modell existiert, sondern der Bedarf. #OpenWeights
Mehr von mir: linkedin.com/in/maurice-putinas
@k-i-news.bsky.socialMicrosoft-Decision-1 gibt keine Texte aus, sondern kalibrierte Wahrscheinlichkeiten pro Antwortoption – für Routing, Klassifikation und Agenten-Steuerung. Basis: Qwen3.5-9B. Entscheiden und Generieren sind offenbar zwei Aufgaben. #OpenWeights
Mehr von mir: linkedin.com/in/maurice-putinas
@prismagenda.bsky.socialMistral just shipped Europe's 1-trillion-parameter "open-weights" model. You can't download it yet.
Tied with GPT-6 Luna on one index, 33rd of 45 on another.
@wideareaai.bsky.socialBy the numbers:
Context Window: 32768
Max Tokens: 8192
Input Cost: 0
Output Cost: 0
#selfhosting #batchinference #gpuoptimization #llm #openweights #inference
@nrouterai.bsky.social1T parameters, open weights, trained in Europe on 3,800 Grace Blackwell GPUs. At ~500GB quantized, Mistral Large 4 isn't a model you run. It's a model you distill from. Open was never about everyone running the flagship. It's about nobody needing permission to build on it.
@spaisee.bsky.socialHirundo’s Qwen release makes political behavior look editable: a merged checkpoint, not a new model trained from scratch. For open-weight developers, that is useful, if the edit does not trade refusals for distortion. #OpenWeights…
@machinelearning.bsky.socialWhy is Mistral's new flagship called Le Chonk? It's 1.05T params (49B active), trained from scratch on 3,800 Grace Blackwell GPUs in Europe. It scores 82% on CyberGym-E2E, where closed models land near zero, partly by refusing. Preview API now, weights due Oct 31. #MachineLearning #OpenWeights
@rpuls.bsky.socialCongrats to Mistral AI on the impressive results and reviews for their new “Le Chonk” open-source / open-weights model 👏🥖🇪🇺
Cant wait to test it.
@hossamudin.bsky.socialشركة Mistral الفرنسية أعلنت أكبر نموذج في تاريخها: Mistral Large 4… بتريليون بارامتر! 🇫🇷🤯
الإعلان كان يوم 6 أكتوبر، واسمه الكودي Le Chonk 🐈 (لقب مرح للنموذج الضخم).
#Mistral #MistralLarge4 #LeChonk #OpenWeights #الذكاء_الاصطناعي #حسام_الدين_حسن #خبير_اونلاين #epreneurs #مهارات_2026
@pulseofnations.lolMistral launched Chonk, a one-trillion-parameter open-weight model trained on its own European compute, staking its case on data sovereignty.
@powerfromspace1.mstdn.social.ap.brid.gy@BijanBowen on YT!
Ling 3.1 Flash First Test – Is THIS Really a #DeepSeek Competitor?
#OpenWeights soon #Ling3.1 #Deepseek #AI
10/6/26
@pulseofnations.lolThe Nvidia-backed startup unveiled Beam, a 501B-parameter open-weight model it says matches GLM-5.2 on reasoning while using 3-4x less inference compute.