“Announcing @MistralAI Small 3.1: multimodal, multilingual, Apache 2.0, the best model in its weight class. 💻 Lightweight: Runs on a single RTX 4090 or a Mac with 32GB RAM, perfect for on-device applications. 🗣️ Fast-Response Conversations: Ideal for virtual assistants and other https://x.com/sophiamyang/status/1901675671815901688
“Introducing the Gemma package, a minimalistic library to use and fine-tune Gemma 🔥 Including docs on: – Fine-tuning – Sharding – LoRA – PEFT – Multimodality – Tokenization !pip install gemma https://x.com/osanseviero/status/1902456220876787763
“Microsoft launched Phi-4-multimodal, a high-performing open weights model capable of processing text, images, and speech simultaneously. Built on a transformer architecture with 5.6 billion parameters, the model delivers impressive results in speech transcription, image https://x.com/DeepLearningAI/status/1902372844421460458
“Mistral have released Mistral Small 3.1, adding image input and a 128k token context window to Mistral Small 3 Key results and info: ➤ @MistralAI Small 3.1 scores an Artificial Analysis Intelligence Index of 35, in line with Mistral 3 and other models such as GPT-4o mini and https://x.com/ArtificialAnlys/status/1902017023917666351
“Too many organizations I speak with make decisions based on the current state of AI, and especially expecting the cost/performance curve to be static. So they build complex systems around the limitations of today’s small, cheaper LLMs. That has been a very bad bet historically. https://x.com/emollick/status/1900341424509341928
Mistral Small 3.1 | Mistral AI https://mistral.ai/news/mistral-small-3-1
“Perplexity’s Sonar API is the best at accurate responses while remaining the most affordable. It scores 91% on SimpleQA while remaining cheaper than even GPT-4o-mini, let alone the expensive GPT-4o. So, bang for the buck – Perplexity is the best. https://x.com/AravSrinivas/status/1902759673549521145
“🚀 Just dropped: SmolDocling-256M brings powerful OCR to your local machine! Only 256M in size but packs full document processing power. https://x.com/fdaudens/status/1901755195505160678
“Nice video on @MistralAI Small 3.1 from @1littlecoder 🫶 https://x.com/sophiamyang/status/1902038297620443612
“icymi few days ago @allen_ai released OLMo-32B, fully open LLM that beats gpt-4o mini and Qwen 2.5 🔥 it’s OLMo aligned with Tülu 3.1 recipe and improvements in RLVR not only that but pre/post-training sets and codebase too 💝 https://x.com/mervenoyann/status/1901961859898458334
“🔥 Mistral-Small-3.1 (24B) is already #3 on @huggingface after just 1 day! Brings SOTA vision + 128k context while fitting on a 32GB MacBook. Crushes benchmarks in reasoning, multilingual & visual tasks. Apache 2.0 licensed 🚀 https://x.com/fdaudens/status/1902111100503572865
“Reka debuted Flash 3, a 21B parameter open-source reasoner model that matches o1-mini —32k context length —Ideal for on-device use —Fine-tuned on synthetic & public data —Powers Nexus, a platform to orchestrate AI workers https://x.com/adcock_brett/status/1901303388651790649
“@MistralAI @huggingface Available on @MistralAI La Plateforme `mistral-small-latest` : https://x.com/sophiamyang/status/1901677125918134276
“Reka AI has launched Reka Flash 3, a new open source 21B parameter reasoning model – the highest scoring model of its size Key details: ➤ Artificial Analysis Intelligence Index of 47, beating almost all non-reasoning models with just 21B total parameters ➤ Stronger than all https://x.com/ArtificialAnlys/status/1902487093273862299
“Mistral AI released Small 3.1, a SOTA multilingual and multimodal LLM —24B (can run on a laptop) —128k token context window —Outperforms Gemma 3 and GPT-4o Mini on most benchmarks —Inference speed of 150 tokens/sec —Open-source under Apache 2.0 license https://x.com/rowancheung/status/1901887637465809285
“Introducing Mistral Small 3.1. Multimodal, Apache 2.0, outperforms Gemma 3 and GPT 4o-mini. https://x.com/MistralAI/status/1901668499832918151
