Vision Language Model

WeatherNext 3 ingests live geostationary satellite mosaics, refreshes hourly, and outputs 5 km forecasts across Search, Gemini, Maps.

Cohere has released Parse (parse-v5.0), a 2.3B-parameter vision language model that converts PDFs, slides and images into Markdown with HTML tables, bounding boxes and image descriptions. It runs at $1.50 per 1,000 pages through the API, or on dedicated Model Vault instances from $2,500 a month. Cohere reports a ParseBench score of 79.2, ahead of Mistral OCR 4, Azure Document Intelligence and Databricks AI Parse — but that figure averages three of the benchmark's five dimensions and drops charts and visual grounding entirely.

This tutorial provides a complete workflow for building a compact, reasoning-focused language model. By streaming the SupraLabs reasoning corpus from Hugging Face, we apply quality filters and curate data for Supervised Fine-Tuning (SFT). Using SmolLM2-135M-Instruct and LoRA, we demonstrate an end-to-end pipeline—from dataset analysis and heuristic cleaning to efficient training and inference—enabling the development of specialized small models without excessive resource requirements

Liquid AI released LFM2.5-VL-3B, a 3.1B-parameter vision-language model built for on-device deployment. It averages 80.7 on ScreenSpot-v2 and lifts RefCOCO grounding from 57.1 to 87.9. Function calling is new to the VL line, with ToolSandbox moving from 26.4 to 59.5. The model fits in roughly 3 GB and decodes 228 tokens/s on an Apple M5 Max.

Dyna Robotics has released Dyna-2, a world-action model pre-trained on more than one million hours of egocentric human video. The technical report establishes three results: a scaling law on human data to 1M hours, the first transfer of that law to unseen robot data, and evidence that video co-training drives cross-embodiment generalization.

Alibaba's Qwen team moved Qwen3.8-Max from preview to general availability, with published per-token pricing and open weights due next week. The 2.4T parameter MoE model accepts text, image and video input across a 1M-token context. No benchmark table has been published.

Onton, a San Francisco-based search and discovery company, has released Ontology 1, a neurosymbolic model for complex, conversational, multimodal product search. On a 90-query...

Google DeepMind has released Gemini Robotics 2, the intelligence layer for its next generation of robots. The release ships three models: a vision-language-action model for whole body humanoid control, Gemini Robotics ER 2 for embodied reasoning and task orchestration, and an on-device VLA that adapts to new robot bodies in hours. One checkpoint drives Apptronik Apollo 2 and a Franka Duo. Only ER 2 is publicly available.

Marktechpost AI has released Token Saver, an open-source MCP extension for Claude Desktop that uses local Hybrid RAG to slash PDF token consumption by up to 99% while ensuring absolute document privacy.

Datalab rewrote Marker as a three-mode pipeline. Version 2 hits 76.0 on olmOCR-bench and sustains 2.9 pages per second on one B200 — over 5× MinerU's pipeline backend, while beating Docling on both accuracy and speed. Here's how it compares against MinerU, Docling and LiteParse, and which one fits your use case.

A single 24GB GPU is the practical floor for serious local inference. This guide compares six open-weight models that fit one card at Q4_K_M. It covers Qwen3.6, Gemma 4, Mistral Small, gpt-oss-20b, and DeepSeek-R1-Distill. Each entry lists VRAM fit, licensing, and the job it does best.

Robbyant, Ant Group's embodied-intelligence unit, has released LingBot-World-Infinity (LingBot-World 2.0). It is a 14B causal video generation model that behaves as an interactive world simulator. The core idea is the Mixture of Bidirectional and Autoregressive (MoBA) attention mask, paired with distribution matching distillation applied over long self-rollout trajectories. Together they target long-horizon drift, the failure mode that smears textures and warps geometry in most interactive world models. A Director-Pilot agentic harness wraps the generator, where a VLM proposes events and the Diffusion Transformer renders them. The report shows a single 60-minute uninterrupted session covering 20 scenarios. But the release is thinner than the paper: one checkpoint, a 480P reference script, no deployment code, no quantitative benchmarks, and a non-commercial CC BY-NC-SA 4.0 license.

Ant Group's Robbyant open-sourced LingBot-Vision, a self-supervised ViT family for dense spatial perception. Masked boundary modeling makes image boundaries a native training signal. The 1B backbone matches or surpasses larger models, and initializes LingBot-Depth 2.0.

Baidu open-sourced Unlimited OCR, a 3B-parameter MoE model that parses dozens of document pages in a single forward pass. Its Reference Sliding Window Attention (R-SWA) holds the KV cache constant, so memory and latency stay flat as output grows. It scores 93.23 on OmniDocBench v1.5, beating the DeepSeek OCR baseline by 6.22 points, under an MIT license.

Zyphra has released Zamba2-VL, a family of open vision-language models at 1.2B, 2.7B, and 7B parameters. The models use a hybrid Mamba2 state-space and Transformer backbone, shipping under Apache 2.0. They stay competitive with comparable Transformer VLMs while cutting time-to-first-token by about an order of magnitude.

StepFun releases Step 3.7 Flash, a 198B MoE model with native vision, 256k context, and Advisor Mode.

Researchers from NVIDIA introduce SANA-WM, an open-source camera-controlled world model that generates 60-second, 720p videos with precise 6-DoF camera control — trained on 64 H100 GPUs and deployable on a single RTX 5090.

Microsoft Research's World-R1 Uses Reinforcement Learning to Force 3D Consistency Into Text-to-Video Models

Meta Reality Labs releases a new foundation model family for human-centric vision that pushes pose estimation, segmentation, and 3D geometry to new state-of-the-art levels — all from a single backbone.

Qwen Team Open-Sources Qwen3.6-35B-A3B: A Sparse MoE Vision-Language Model with 3B Active Parameters and Agentic Coding Capabilities

Liquid AI just released LFM2.5-VL-450M, an updated version of its earlier LFM2-VL-450M vision-language model. The new release introduces bounding box prediction, improved instruction following,...

Running powerful AI on your smartphone isn't just a hardware problem — it's a model architecture problem. Most state-of-the-art vision encoders are enormous, and...

In the current landscape of computer vision, the standard operating procedure involves a modular 'Lego-brick' approach: a pre-trained vision encoder for feature extraction paired...

IBM has announced the release of Granite 4.0 3B Vision, a vision-language model (VLM) engineered specifically for enterprise-grade document data extraction. Departing from the...

Microsoft has released Phi-4-reasoning-vision-15B, a 15 billion parameter open-weight multimodal reasoning model designed for image and text tasks that require both perception and selective...

Current end-to-end robotic policies, specifically Vision-Language-Action (VLA) models, typically operate on a single observation or a very short history. This 'lack of memory' makes...

In the escalating 'race of "smaller, faster, cheaper' AI, Google just dropped a heavy-hitting payload. The tech giant officially unveiled Nano-Banana 2 (technically designated...

Building simulators for robots has been a long term challenge. Traditional engines require manual coding of physics and perfect 3D models. NVIDIA is changing...

Frontier multimodal models usually process an image in a single pass. If they miss a serial number on a chip or a small symbol...

How do you build a single vision language action model that can control many different dual arm robots in the real world? LingBot-VLA is Ant...

Liquid AI has introduced LFM2.5, a new generation of small foundation models built on the LFM2 architecture and focused at on device and edge...

Zhipu AI has open sourced the GLM-4.6V series as a pair of vision language models that treat images, video and tools as first class...

Jina AI has released Jina-VLM, a 2.4B parameter vision language model that targets multilingual visual question answering and document understanding on constrained hardware. The...

Tencent Hunyuan has released HunyuanOCR, a 1B parameter vision language model that is specialized for OCR and document understanding. The model is built on...

How do you build one open model that can reliably understand text, images, audio and video while still running efficiently? A team of researchers...

How can we get large model level multimodal reasoning for documents, charts and videos while running only a 3B class model in production? Baidu...

Liquid AI released LFM2-VL-3B, a 3B parameter vision language model for image text to text tasks. It extends the LFM2-VL family beyond the 450M...

How do you convert complex, multilingual documents—dense layouts, small scripts, formulas, charts, and handwriting—into faithful structured Markdown/JSON with state-of-the-art accuracy while keeping inference latency...

Do you actually need a giant VLM when dense Qwen3-VL 4B/8B (Instruct/Thinking) with FP8 runs in low VRAM yet retains 256K→1M context and the...

IBM has released Granite-Docling-258M, an open-source (Apache-2.0) vision-language model designed specifically for end-to-end document conversion. The model targets layout-faithful extraction—tables, code, equations, lists, captions,...

In the domain of multimodal AI, instruction-based image editing models are transforming how users interact with visual content. Just released in August 2025 by...

Zhipu AI has officially released and open-sourced GLM-4.5V, a next-generation vision-language model (VLM) that significantly advances the state of open multimodal AI. Based on...

Tencent's Hunyuan team has introduced Hunyuan-A13B, a new open-source large language model built on a sparse Mixture-of-Experts (MoE) architecture. While the model consists of...

The Alibaba Qwen team has introduced Qwen-VLo, a new addition to its Qwen model family, designed to unify multimodal understanding and generation within a...

The DeepSeek Researchers just released a super cool personal project named 'nano-vLLM', a minimalistic and efficient implementation of the vLLM (virtual Large Language Model)...

In recent years, contrastive language-image models such as CLIP have established themselves as a default choice for learning vision representations, particularly in multimodal applications...

In recent years, vision-language models (VLMs) have advanced significantly in bridging image, video, and textual modalities. Yet, a persistent limitation remains: the inability to...

OpenAI’s GPT-4o represents a new milestone in multimodal AI: a single model capable of generating fluent text and high-quality images in the same output...

Modern vision-language models have transformed how we process visual data, yet they often fall short when it comes to fine-grained localization and dense feature...

Vision‐language models (VLMs) have long promised to bridge the gap between image understanding and natural language processing. Yet, practical challenges persist. Traditional VLMs often...

The integration of visual and textual data in artificial intelligence presents a complex challenge. Traditional models often struggle to interpret structured visual documents such...

Vision-Language Models (VLMs) have significantly expanded AI’s ability to process multimodal information, yet they face persistent challenges. Proprietary models such as GPT-4V and Gemini-1.5-Pro...

In the evolving landscape of artificial intelligence, integrating vision and language capabilities remains a complex challenge. Traditional models often struggle with tasks requiring a...

Multimodal reasoning—the ability to process and integrate information from diverse data sources such as text, images, and video—remains a demanding area of research in...

While multimodal models (LMMs) have advanced significantly for text and image tasks, video-based models remain underdeveloped. Videos are inherently complex, combining spatial and temporal...

The rapid development of artificial intelligence (AI) has produced models with powerful capabilities, such as language understanding and vision processing. However, deploying these models...

In recent years, there has been a growing demand for machine learning models capable of handling visual and language tasks effectively, without relying on...

Edge AI has long faced the challenge of balancing efficiency and effectiveness. Deploying Vision Language Models (VLMs) on edge devices is difficult due to...

Multimodal models represent a significant advancement in artificial intelligence by enabling systems to process and understand data from multiple sources, like text and images....

The release of Pixtral 12B by Mistral AI represents a groundbreaking leap in the multimodal large language model powered by an impressive 12 billion...

Large Language Models (LLMs), initially limited to text-based processing, faced significant challenges in comprehending visual data. This limitation led to the development of Visual...

Researchers at Alibaba have announced the release of Qwen2-VL, the latest iteration of vision language models based on Qwen2 within the Qwen model family....

Multimodal large language models (MLLMs) represent a significant leap in artificial intelligence by combining visual and linguistic information to understand better and interpret complex...