Datalab has launched lift, a 9B open-weights imaginative and prescient mannequin for structured extraction. You go it a JSON schema, and it returns a JSON object that matches. The mannequin reads PDFs and pictures straight, then decodes towards your schema. This is Datalab’s first mannequin constructed purely for extraction. The staff already ships open-source OCR…
MoonMath AI group has launched a bf16 ahead consideration kernel for AMD’s MI300X GPU. It is written in HIP, not hand-written meeting. The code is open-source below the MIT license. The MoonMath.ai group studies it beats AITER v3, AMD’s personal optimized kernel, on each examined form. Bare-metal entry got here from HotAisle, an AMD cloud…
Getting prompts proper remains to be the toughest a part of delivery dependable LLM functions. Small wording modifications can swing accuracy by 20 %. What works on just a few examples usually breaks at scale. When a multi-step pipeline returns a incorrect reply, discovering the failing step means inspecting intermediate outputs by hand. Cisco AI…
TLDR YaFF is Yandex’s open-source zero-copy wire format for Protobuf — Apache 2.0, at the moment C++, v0.1.0. The .proto file stays the supply of reality; solely the bodily reminiscence format adjustments. On Yandex’s benchmarks, the Flat Layout reads scorching knowledge ~3.8× sooner than FlatBuffers, inside 1.2× of a uncooked C++ struct. Four layouts —…
While current breakthroughs in AI reasoning have largely been pushed by large scale, pouring in billions of parameters to cross complicated cognitive thresholds—VibeThinker-3B is charting a very completely different path. Created by researchers from Sina Weibo Inc (China), this 3-billion-parameter mannequin proves that effectivity can punch far above its weight class. Released underneath an open-source…
This week, Liquid AI launched two new retrieval fashions. They are (*11*) and LFM2.5-Embedding-350M. Both maintain 350M parameters. Both are the primary bidirectional members of the LFM household. They construct on LFM2.5-350M-Base, launched in March. The pair targets quick multilingual and cross-lingual search throughout 11 languages. Their footprint is sufficiently small to run virtually wherever….
The Qwen workforce has launched three embodied AI fashions, grouped as Qwen-Robot-Suite. The three are Qwen-RobotManip, Qwen-RobotWorld, and Qwen-RobotNav. Each is constructed on a Qwen vision-language spine and targets a distinct robotics downside. Qwen-RobotManip is a Vision-Language-Action mannequin for manipulation, constructed on Qwen3.5-4B. Qwen-RobotWorld is a language-conditioned video world mannequin with a 60-layer MMDiT and…
Tokyo-based Sakana AI shipped its first industrial product ‘Sakana Marlin’ this week. Sakana staff positions it as a Virtual CSO (Chief Strategy Officer). It is a B2B autonomous analysis agent constructed for enterprises. Marlin doesn’t reply in seconds like a chatbot. You give it one analysis matter. It then runs autonomously for up to about…
Zyphra has launched Zamba2-VL, a household of open vision-language fashions. The launch covers three sizes: 1.2B, 2.7B, and 7B parameters. Each mannequin is constructed on the Zamba2 hybrid SSM–Transformer spine. Vision-language fashions (VLMs) learn photos and textual content collectively. They reply questions on charts, paperwork, and pictures. Most open VLMs use a dense Transformer because…
Google AI group together with the Google DeepMind researchers have simply launched DiffusionGemma, an experimental open mannequin for textual content era. It makes use of textual content diffusion as an alternative of normal autoregressive decoding. The mannequin ships beneath a permissive Apache 2.0 license. Google positions it for devs and researchers exploring speed-critical, interactive native…