Model Releases: 310 curated resources
Curated material on model releases within frontier.
310 resources · page 10 of 10
Flow Matching
Creator not supplied; see original source.
arXiv paper (2022) "Flow Matching for Generative Modeling", introducing flow matching for training continuous normalising flows.
GANs (Goodfellow)
Creator not supplied; see original source.
arXiv paper (2014) by Goodfellow et al. introducing generative adversarial networks (GANs) and adversarial training.
Gemini image generation got it wrong (postmortem)
Creator not supplied; see original source.
Google post explaining what went wrong with Gemini's image generation of people and the planned fixes.
Hi3D 产品概述
Creator not supplied; see original source.
Hi3D API overview in Chinese: Math Magic's 3D generation service that turns 2D images into 3D models.
Hi3D API introduction
Creator not supplied; see original source.
Hi3D API introduction: image-to-3D generation API from Math Magic based on its Sparc3D and Ultra3D models.
Krea 2 Technical Report
Creator not supplied; see original source.
Krea 2 technical report on its open-weight text-to-image foundation models, including architecture and training details.
Latent Diffusion (Stable Diffusion)
Creator not supplied; see original source.
arXiv paper (2021) introducing latent diffusion models, the basis of Stable Diffusion.
Meshy API documentation and changelog
Creator not supplied; see original source.
Meshy API changelog listing feature updates, model versions and migration notes for its 3D generation API.
NeRF
Creator not supplied; see original source.
arXiv paper (2020) "NeRF: Representing Scenes as Neural Radiance Fields for View Synthesis", introducing neural radiance fields (NeRF) for novel view synthesis.
Nerfstudio documentation
Creator not supplied; see original source.
Nerfstudio documentation: modular framework for creating, training and viewing neural radiance fields (NeRFs).
Qwen3 Technical Report
Creator not supplied; see original source.
arXiv paper (2025) Qwen3 Technical Report describing the Qwen3 model family, architecture and training.
Rodin Gen-2.5 中文 API specification
Creator not supplied; see original source.
Hyper3D Rodin Gen-2.5 API specification in Chinese: request parameters, authentication and responses for 3D asset generation.
Rodin official generation skill library
Creator not supplied; see original source.
Deemos rodin3d-skills GitHub repo: official agent skills for Rodin 3D generation.
Sloyd API getting started
Creator not supplied; see original source.
Sloyd API getting-started guide: REST API for generating 3D models from text and images, with create-poll-download examples.
Stability 3D
Creator not supplied; see original source.
Stability AI page on Stable Zero123, a model for generating 3D objects and novel views from a single image.
Tripo developer documentation
Creator not supplied; see original source.
Tripo developer documentation: API for generating 3D models from text and images, with authentication and task handling.
VAE (Kingma & Welling)
Creator not supplied; see original source.
arXiv paper (2013) "Auto-Encoding Variational Bayes", introducing the variational autoencoder (VAE).
VARCO 3D 2.0 release documentation
Creator not supplied; see original source.
VARCO 3D 2.0 release notes from NCSOFT describing upgrades to its 3D generation model and benchmark results.
VARCO 3D workflow and MCP guide
Creator not supplied; see original source.
VARCO 3D release note on connecting generation, rigging and animation in custom workflows and agent tools.
Whisper paper
Creator not supplied; see original source.
arXiv paper (2022) introducing Whisper, speech recognition trained on large-scale weakly supervised audio.
GEMA v OpenAI decision (MediaLaws)
Creator not supplied; see original source.
MediaLaws analysis of the Munich Regional Court decision in GEMA v. OpenAI on song lyrics and model training.
Hitem3D 2.0: Multi-View Guided Native 3D Texture Generation
Creator not supplied; see original source.
arXiv paper (2026) describing Hitem3D 2.0, multi-view guided native 3D texture generation.