SuperAPI AI Directory

All AI tools

35,738 results · Page 323 of 993

OnlineImageUpscaler

AI images

OnlineImageUpscaler is a free online AI image upscaling tool that can boost the resolution of JPG, PNG, and WEBP images to 2K, 4K, 8K, or even 16K in one click (login required and subject to credit limits), while also enhancing clarity, sharpening edges, removing blur, restoring details, and even colorizing black-and-white old photos. No design experience or complex settings are required, and registered users can batch process up to 20 images. It is suitable for scenarios such as professional photography, e-commerce product images, large-format printing, social media content, real estate and architectural displays, and app and website graphics.

Read more

ontheia

AI developer platforms

Ontheia is an open-source, self-hosted AI agent platform that emphasizes data sovereignty and compliance. It supports multiple model providers (Claude, GPT, Grok, Gemini, Ollama, and any OpenAI-compatible model), multi-user and role-based access control, and natively connects to external tools based on the open Model Context Protocol (MCP). The platform includes built-in Agent Skills reusable modules, Cron scheduled tasks, a visual workflow engine, long-term memory based on pgvector (RAG), and inter-agent delegation and collaboration. Data always stays on your own servers, achieving GDPR compliance by design through PostgreSQL row-level security. At least one AI provider API key is required for conversations, and memory features additionally require a provider that supports embeddings.

Read more

Open challenges in LLM research

AI developer platforms

This is a review article written by Huyenchip on the open challenges in large language model (LLM) research, summarizing 10 major research directions in the current LLM field, including reducing hallucinations, optimizing context length, multimodal fusion, improving speed and reducing cost, designing new architectures, developing GPU alternatives, making agents usable, improving human preference learning, and more. Based on the author's conversations with people in industry and academia, the article aims to provide directional reference for researchers and practitioners.

Read more

open-dreamer

AI developer platforms

Open Dreamer is an open-source JAX/Flax (NNX) implementation of the Dreamer 4 world model, focused on training and running a world model in real time on Minecraft/VPT-style game data. The repository provides the complete training pipeline code: including a causal video tokenizer, an action-conditioned latent dynamics model, and tools for generating rollouts and computing the FVD quality metric. Users can train the tokenizer, tokenize a Minecraft/VPT-style MP4 dataset, train the latent dynamics model, generate videos, and evaluate FVD. The project also provides a real-time demo in the browser, where you can enter a generated Minecraft world and, via the Game ⟷ Dream toggle, transition frame by frame between real gameplay footage and model-generated footage.

Read more

open-gpt.app

AI search

open-gpt.app is a platform for a collection of ChatGPT apps where users can use a variety of ready-made ChatGPT apps in seconds or create their own apps. The platform provides a rich set of app examples covering fields such as entertainment, writing, programming, and daily life, such as Pokémon data lookup, movie recommendations, regex generation, Midjourney prompt generation, grammar correction, summary generation, startup plan inspiration, resume editing, blog writing, Xiaohongshu copy generation, weekly report generation, tarot divination, and more. Users can quickly invoke these apps without programming to meet diverse text-processing needs.

Read more

open-managed-agents

AI developer platforms

Open Managed Agents (OpenMA) is an open-source, self-hosted alternative to Claude Managed Agents and the OpenAI Agents API, maintained by openma-ai on GitHub. It is used to run agents with persistent sessions, sandboxed tools, memory, encrypted credentials, and crash recovery on your own infrastructure. Users can self-host with Node.js or Docker, deploy to Cloudflare, or use the hosted version of OpenMA (app.openma.dev) directly. The platform supports bring-your-own model keys (BYOK), allowing free choice of model, harness, and sandbox, and provides Claude-compatible /v1/agents and /v1/sessions APIs as well as an OpenAI Agents API-compatible /openai/v1 endpoint, along with a Console UI.

Read more

Open-Qwen2VL

AI developer platforms

Open-Qwen2VL is the official code repository for an efficiently computed pre-trained, fully open-source multimodal large language model (MLLM) aimed at academic resource environments, and the paper has been accepted by COLM 2025. The repository supports the complete pipeline from data quality scoring, high-quality data selection and re-sharding, multimodal sequence packing, to large-scale image-text data pre-training, small-scale and large-scale supervised fine-tuning, and multimodal benchmark evaluation. The project has released pre-trained model and instruction fine-tuned model checkpoints, pre-training data in webdataset format, and a technical report, and supports packing sequences into a single JSON file containing base64 images and text to save disk space.

Read more