{"company":{"name":"Mistral AI","slug":"mistral-ai","website":"https://mistral.ai","category":"ai-ml"},"question":"What has Mistral AI shipped recently?","answer":"In the last 30 days, Mistral AI shipped 17 tracked updates. The most recent was \"Facciamo avanzare la frontiera aperta.\" on 2026-10-16.","window":{"days":30,"updateCount":17,"returned":17},"generatedAt":"2026-10-16T20:23:16.223Z","updates":[{"title":"Facciamo avanzare la frontiera aperta.","summary":"Mistral AI unveiled several new and updated open-weight models, including Mistral Large 4 (49B active MoE), Mistral Medium 3.5 (128B), Mistral Small 4 (6.5B active), OCR 4.1 (document extraction), Voxtral TTS (text-to-speech), and specialized models for robotics, physics, and safety. These models unify instruction, reasoning, and agentic capabilities with multimodal input.","date":"2026-10-16","dateIsEstimated":true,"signalType":"product_launch","signalTypeLabel":"Launch","sourceUrl":"https://mistral.ai/it/research/","publisher":"mistral.ai"},{"title":"Introducing Timestral","summary":"Mistral AI unveiled Timestral, its first family of time-series foundation models, with four sizes (5M to 209M parameters) delivering state-of-the-art probabilistic, multivariate forecasts. The largest model (209M params) tops public benchmarks while outperforming larger competitors, and all models run efficiently on CPUs with zero-shot forecasting capabilities.","date":"2026-10-15","dateIsEstimated":false,"signalType":"product_launch","signalTypeLabel":"Launch","sourceUrl":"https://mistral.ai/news/timestral/","publisher":"mistral.ai"},{"title":"New Material joins Mistral to design the frontier","summary":"Mistral AI announced the acquisition of New Material, a design-first AI consultancy, to establish a new Solutions Design team. The move aims to embed human-centered design into Mistral’s enterprise and government AI solutions, accelerating sovereign AI delivery with tighter customer feedback loops.","date":"2026-10-14","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://newmaterial.ai/stories/newmaterial-joins-mistral","publisher":"newmaterial.ai"},{"title":"Synapse Medicine","summary":"Synapse Medicine, used by over 30% of French GPs and 300 hospitals, integrates Mistral’s AI models to deliver personalized, evidence-based medical recommendations tailored to patient clinical contexts. This integration enables faster, more informed physician decision-making and improved patient outcomes.","date":"2026-10-06","dateIsEstimated":true,"signalType":"case_study","signalTypeLabel":"Case Study","sourceUrl":"https://mistral.ai/customers/synapse-medicine/","publisher":"mistral.ai"},{"title":"Mistral AI on X: \"Meet Mistral Large 4, aka Le Chonk. \n\n• 1T parameters, natively multimodal. 49B active.\nIt is the best open weights model from US or Europe on aggregated benchmarks.\n• State-of-the-art on critical workloads, including cyber defense, manufacturing and finance and it surpasses closed frontier models on visual grounding.\n• Forged in Europe end-to-end and is deployable from Europe via our own Mistral Cloud infrastructure.\n• Available to all via API today. Working with cybersecurity partners privately.\n\nOpen weights release end of October.\"","summary":"Mistral AI unveiled Mistral Large 4, codenamed Le Chonk, a 1 trillion-parameter natively multimodal model with 49 billion active parameters. It claims to be the best open-weights model from the US or Europe on aggregated benchmarks, excelling in cyber defense, manufacturing, finance, and visual grounding. Deployable via Mistral Cloud in Europe, it is available via API today with an open-weights release planned for late October.","date":"2026-10-06","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://x.com/MistralAI/status/2107457414387622310?launch_app_store=true&ct=engagement_view_post","publisher":"x.com"},{"title":"Introducing Mistral Large 4 | Mistral","summary":"Mistral AI unveiled Mistral Large 4 (ML4), a 1 trillion-parameter multimodal model with 49 billion active parameters, offering state-of-the-art performance in coding, cybersecurity, finance, and multimodal understanding. The model is available via a public preview API and will release weights by October 31, 2026.","date":"2026-10-06","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://mistral.ai/news/mistral-large-4/","publisher":"mistral.ai"},{"title":"Mistral Large 4","summary":"Mistral AI introduced Mistral Large 4, a new open-weight multimodal model featuring a Mixture-of-Experts architecture with 52B active and 1.05T total parameters, including a 1.6B vision encoder. It supports structured outputs, function calling, document Q&A, and batching, with pricing reduced by 50% for input, cached input, and output tokens.","date":"2026-10-06","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://docs.mistral.ai/models/mistral-large-4-0","publisher":"docs.mistral.ai"},{"title":"Présentation de Mistral Large 4 | Mistral","summary":"Mistral AI unveiled Mistral Large 4 (ML4), a 1-trillion-parameter multimodal model with 49B active parameters, now in public preview via Mistral Studio. Weights will be released by month-end. ML4 leads in coding, agentic workflows, and multimodal tasks, outperforming open-weight models globally and rivaling closed models in areas like visual grounding and cybersecurity.","date":"2026-10-06","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://mistral.ai/fr/news/mistral-large-4/","publisher":"mistral.ai"},{"title":"Mistral AI on X: \"Mistral Large 4 is one of the world's strongest AI models for cybersecurity.\"","summary":"Mistral AI announced Mistral Large 4, positioning it as one of the world's strongest AI models for cybersecurity. Demonstrations highlight its performance in solving 18 of 19 CTF challenges and reverse-engineering malware to identify Cobalt Strike samples and extract IoCs.","date":"2026-10-06","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://x.com/MistralAI/status/2107532329937723861?launch_app_store=true&ct=engagement_view_post","publisher":"x.com"},{"title":"Presentazione di Mistral Large 4 | Mistral","summary":"Mistral AI unveiled Mistral Large 4 (ML4), a 1-trillion-parameter natively multimodal model with 49 billion active parameters, now available in public preview via Mistral Studio API. The model excels in coding, agentic workflows, cybersecurity, finance, and legal tasks, outperforming other open-weight models globally and rivaling closed models in areas like visual grounding.","date":"2026-10-06","dateIsEstimated":false,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://mistral.ai/it/news/mistral-large-4/","publisher":"mistral.ai"},{"title":"Il cloud per l'IA di frontiera.","summary":"Mistral AI introduced Mistral Compute, a dedicated GPU cloud platform optimized for large-scale AI training and inference. The service offers managed Slurm and Kubernetes clusters, direct hardware access, built-in observability, and enterprise-grade security with BYOK encryption. It targets European enterprises seeking sovereign AI infrastructure.","date":"2026-10-06","dateIsEstimated":true,"signalType":"product_launch","signalTypeLabel":"Launch","sourceUrl":"https://mistral.ai/it/cloud/compute/","publisher":"mistral.ai"},{"title":"Inferenza per costruire la tua avanguardia in Europe.","summary":"Mistral AI introduced regional inference endpoints in the EU and US for supported models, allowing customers to choose data residency and reduce latency with a 10% surcharge. New pricing includes a 50% discount for batch processing and up to 90% savings on cached input tokens. Enterprise APIs now offer regional controls, system-level SLAs, and higher rate limits.","date":"2026-10-06","dateIsEstimated":true,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://mistral.ai/it/cloud/inference/","publisher":"mistral.ai"},{"title":"L'inférence pour construire votre IA de pointe en Europe.","summary":"Mistral AI introduced regional inference endpoints in the EU and US for supported models, allowing customers to process requests in their preferred region for data residency and latency control. New pricing tiers include a 50% discount for batch processing and up to 90% savings on repeated input tokens via caching. Enterprise APIs now offer regional data controls, system-level SLAs, and premium support.","date":"2026-10-06","dateIsEstimated":true,"signalType":"feature_update","signalTypeLabel":"Feature","sourceUrl":"https://mistral.ai/fr/cloud/inference/","publisher":"mistral.ai"},{"title":"IA personalizzata per il tuo dominio. Sviluppata insieme.","summary":"Mistral AI unveiled partnerships with ASML to innovate semiconductor AI solutions, Helsing to develop Vision-Language-Action models for European defense, and HTX to fine-tune models for Singapore’s Home Team. These collaborations focus on domain-specific AI tailored to critical industries and public sector needs.","date":"2026-10-04","dateIsEstimated":true,"signalType":"partnership","signalTypeLabel":"Partnership","sourceUrl":"https://mistral.ai/it/solutions/custom-model-training/","publisher":"mistral.ai"},{"title":"Une intelligence adaptée à votre domaine. Développée en collaboration.","summary":"Mistral AI introduced a custom model training service enabling enterprises to fine-tune or pre-train LLMs using proprietary data for domain-specific use cases like finance, healthcare, and industrial production. The offering emphasizes sovereign data control, flexible deployment (on-prem, cloud, edge), and continuous model reinforcement.","date":"2026-10-04","dateIsEstimated":true,"signalType":"product_launch","signalTypeLabel":"Launch","sourceUrl":"https://mistral.ai/fr/solutions/custom-model-training/","publisher":"mistral.ai"},{"title":"Models Overview","summary":"Mistral AI introduced several new and updated models, including OCR 4.1 for document understanding, Codestral for code generation, and Voxtral Mini Transcribe 2 and Realtime for transcription. Mistral Medium 3.5 and Mistral Small 4 were highlighted as frontier-class and hybrid models, respectively.","date":"2026-10-04","dateIsEstimated":true,"signalType":null,"signalTypeLabel":null,"sourceUrl":"https://docs.mistral.ai/models","publisher":"docs.mistral.ai"},{"title":"Hallo, Deutschland!","summary":"Mistral AI launched a new hub in Munich to drive industrial AI innovation in Germany, Europe’s largest economy. The hub will focus on Physics AI and Industrial AI, partnering with BMW, Siemens Energy, and TUM to solve heavy industry challenges like crash simulations and aerodynamic digital twins.","date":"2026-09-28","dateIsEstimated":false,"signalType":"expansion","signalTypeLabel":"Expansion","sourceUrl":"https://mistral.ai/news/hallo-deutschland/","publisher":"mistral.ai"}],"attribution":{"source":"Spyingbee","url":"https://spyingbee.com/updates/mistral-ai","citation":"Spyingbee, \"Mistral AI updates\", https://spyingbee.com/updates/mistral-ai (retrieved 2026-10-16)"}}