<h1 id="qwen3-vl">Qwen3-VL</h1>Meet Qwen3-VL — the most powerful vision-language model in the Qwen series to date.This generation delivers comprehensive upgrades across the board: superior text understanding &amp; generation, deeper visual perception &amp; reasoning, extended context length, enhanced spatial and video dynamics comprehension, and stronger agent interaction capabilities.Available in Dense and MoE architectures that scale from edge to cloud, with Instruct and reasoning‑enhanced Thinking editions for flexible, on‑demand deployment.Key Enhancements:<ul><li>Visual Agent: Operates PC/mobile GUIs—recognizes elements, understands functions, invokes tools, completes tasks.</li><li>Visual Coding Boost: Generates Draw.io/HTML/CSS/JS from images/videos.</li><li>Advanced Spatial Perception: Judges object positions, viewpoints, and occlusions; provides stronger 2D grounding and enables 3D grounding for spatial reasoning and embodied AI.</li><li>Long Context &amp; Video Understanding: Native 256K context, expandable to 1M; handles books and hours-long video with full recall and second-level indexing.</li><li>Enhanced Multimodal Reasoning: Excels in STEM/Math—causal analysis and logical, evidence-based answers.</li><li>Upgraded Visual Recognition: Broader, higher-quality pretraining is able to “recognize everything”—celebrities, anime, products, landmarks, flora/fauna, etc.</li><li>Expanded OCR: Supports 32 languages (up from 19); robust in low light, blur, and tilt; better with rare/ancient characters and jargon; improved long-document structure parsing.</li><li>Text Understanding on par with pure LLMs: Seamless text–vision fusion for lossless, unified comprehension.</li></ul>Model Architecture Updates:<img src="https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen3-VL/qwen3vl_arc.jpg" width="80%" /><ol><li>Interleaved-MRoPE: Full‑frequency allocation over time, width, and height via robust positional embeddings, enhancing long‑horizon video reasoning.</li><li>DeepStack: Fuses multi‑level ViT features to capture fine‑grained details and sharpen image–text alignment.</li><li>Text–Timestamp Alignment: Moves beyond T‑RoPE to precise, timestamp‑grounded event localization for stronger video temporal modeling.</li></ol>

Qwen3-VL

qwen3VL_8b.safetensors

Ideogram 4 is a foundation model trained entirely from scratch, not a fine-tune or distillation of any existing checkpoint. It is a flow-matching text-to-image model built on a fully single-stream Diffusion Transformer (DiT) architecture.Architecture:<ul><li>Fully single-stream DiT. Text and image tokens are concatenated into one unified sequence and processed through the same 34-layer transformer, with no separate text or image branches. This enables deep cross-modal interaction at every layer.</li><li>Vision-language model as text encoder. Instead of a text-only encoder like CLIP or T5, Ideogram 4 uses <a target="_blank" rel="ugc" href="https://huggingface.co/Qwen/Qwen3-VL-8B-Instruct">Qwen3-VL-8B-Instruct</a>, a full vision-language model that provides far richer understanding of visual concepts. Hidden states are extracted from 13 intermediate layers and concatenated, giving the model multi-scale semantic features ranging from surface-level token information to deep compositional understanding.</li><li>Dual-branch classifier-free guidance. The conditional (positive) and unconditional (negative) branches can be independently refined, enabling separate control over prompt adherence and image quality.</li><li>Flexible resolution. Native support for any resolution from 256 to 2048 (multiples of 16), with aspect ratios up to 6:1. A single model handles everything from square thumbnails to ultrawide banners, with the noise schedule auto-adjusting per resolution.</li></ul>Key Capabilities:<ul><li>Extreme controllability. Ideogram 4 is trained on structured JSON captions, giving users unprecedented control over composition, style, lighting, color palette, typography, and spatial layout, all from a single prompt.</li><li>State-of-the-art text rendering. Ideogram 4 delivers best-in-class in-image text generation (signage, logos, captions, watermarks, multi-line text) with high fidelity directly from the prompt.</li><li>Spatial layout control. Bounding-box coordinates in the prompt allow explicit placement of subjects, text elements, and background regions.</li><li>Color palette conditioning. Specify hex colors in the prompt to steer the image's dominant color scheme.</li></ul>For full architecture details, see <a target="_blank" rel="ugc" href="https://github.com/ideogram-oss/ideogram4/blob/main/docs/model_architecture.md">docs/model_architecture.md</a>. For a walkthrough of how the pipeline components fit together, see <a target="_blank" rel="ugc" href="https://github.com/ideogram-oss/ideogram4/blob/main/docs/pipeline.md">docs/pipeline.md</a>.<h2 id="prompting-guide">Prompting Guide</h2>Ideogram 4 is trained exclusively on structured JSON captions. While plain-text prompts work, you will get the best results by providing a JSON object that follows our caption schema.Key points:<ul><li>Use JSON prompts for maximum controllability — the model was trained on them and understands the structure natively.</li><li>Color palette conditioning — specify a <code>colour_palette</code> array of hex colors in the style description to steer the image's color scheme.</li><li>Aspect ratio flexibility — Ideogram 4 supports a wide range of aspect ratios (any multiple-of-16 resolution from 256 to 2048 on each side). This is a key advantage for practical use: portraits, landscapes, banners, phone wallpapers, social media formats, etc.</li><li>Bounding-box layout — specify <code>bbox</code> coordinates in the prompt to explicitly place subjects, text elements, and background regions.</li><li>Compositional control — use <code>compositional_deconstruction</code> with bounding boxes and per-element descriptions for precise spatial layout.</li></ul>Why JSON-only training? We train exclusively on JSON so that training and inference share a single, common prompt format. The training captions themselves are deliberately extremely descriptive: each JSON exhaustively describes everything in the image to maximize training efficiency. The more text-to-image relationships each caption pins down, the more grounded supervision the model extracts from a single training pair, rather than having to infer those relationships across many sparsely-captioned samples.Why JSON at inference time? Because the model was trained on captions that name every object explicitly, the most reliable way to get every requested object rendered is to mirror that pattern. Plain-text prompts still work, but won't perform as well since the model was only trained on structured JSON captions.Don't want to write JSON by hand? That's what magic prompt is for: it uses an LLM to expand a plain-text prompt into a full structured caption before generation, so you get JSON-quality results from a casual prompt. It runs by default in <code>run_inference.py</code> (see the <a target="_blank" rel="ugc" href="https://huggingface.co/ideogram-ai/ideogram-4-fp8#cli">CLI</a> section).See <a target="_blank" rel="ugc" href="https://github.com/ideogram-oss/ideogram4/blob/main/docs/prompting.md">docs/prompting.md</a> for a full guide.

Files with SHA256: e462e9e0c3b9313ae17f82040d7c77beb92d7aef3e40692d7803228dab7c3b98