MiniMax releases open-weight 4K image model
MiniMax has released H3-Image, an open-weight model that it says generates 4096 × 4096 images in under 10 seconds, so high-resolution output becomes practical for everyday creative work.
MiniMax announced H3-Image, an open-weight image generation model built on its H3 model. It handles text-to-image generation, reference-based generation and image editing.
The company says H3-Image can render a full 4K (4096 × 4096, 16MP) image in under 10 seconds, which it describes as up to 5× faster than "mainstream alternatives". The model accepts up to 16 reference images. MiniMax claims it has state-of-the-art prompt adherence among open-weight models. The post doesn't name the competing models, benchmarks or hardware behind these comparisons.
MiniMax says no single change produced the gains. Instead, it points to improvements across the post-training pipeline built on top of H3's pre-training: data, reinforcement learning, sampling-step reduction, prompt engineering and deployment optimization. The company also says the model avoids the overly smooth look common in AI-generated images, though it gives no quantitative quality results.
The weights are on Hugging Face. MiniMax says it hopes the release will help narrow the gap between open-weight and proprietary image models. The model is also available through the MiniMax API Platform, MiniMax Code and Hailuo AI.
On the API, Standard mode runs 1.5–2× as fast as mainstream alternatives and Highspeed mode runs 3–4× as fast, according to MiniMax. Pricing is 50–70% of comparable offerings. For launch week, the company is cutting the standard price in half for requests at 4K resolution and above.
The announcement does not say what license the weights use, how large the model is, or what hardware is needed to run it locally.