Loading...
Claude Mythos 5 is Anthropic’s trusted-access model for cybersecurity and biology research. It is described as state of the art in cybersecurity, biology research, and healthcare, with pricing published separately from standard Claude plans.
NVIDIA Nemotron 3 Ultra is an open model for long-running agent workflows that need strong reasoning, tool use, and efficient multi-turn execution. It is positioned for developers building complex agentic systems that require high accuracy, high throughput, and open deployment options.
Gemma 4 12B is Google’s mid-sized multimodal model for developers who want local, laptop-class inference with vision and audio support. It is available under Apache 2.0 and can be used with local tools, downloadable checkpoints, and Google Cloud deployment paths.
EchoFlow is an Android chat app for OpenRouter that uses your own API key, stores conversations locally on your phone, and supports switching between models mid-chat. It is free, requires no account, and keeps chat history off the server.
Tokenwise is an LLM observability and cost optimization proxy for production apps. It helps teams monitor usage, detect waste, and apply guarded cost-saving changes without rewriting their SDK integration.
MiniCPM5-1B is a compact open-source language model on Hugging Face for local assistants, coding agents, tool use, and reasoning. It includes a 131,072-token context window, Think and No Think chat modes, and runtime-specific releases such as GGUF and MLX.
Command A+ is Cohere’s open-source enterprise language model for agentic workflows, offering multimodal, multilingual, and tool-use capabilities with private deployment options. It is aimed at teams building production AI systems that need long context and efficient inference.
MashuPack is a browser-based tool for turning selected parts of a local repository into one structured text file for ChatGPT, Claude, and similar AI chats. It helps developers package code context without uploading the repo or copying files manually.
Krater is an AI workspace that combines 350+ models, multimodal creation tools, and scheduled workflows in one subscription. It’s aimed at people and teams that want to manage chat, media generation, document analysis, and app-connected tasks from a single product.
Harbor is an open-source CLI and companion app for running a pre-wired local AI stack. It helps users start and manage local LLM services, supporting tools, and related workflows with minimal manual setup.
Perceptron Mk1 is a closed-source vision model for video understanding and embodied reasoning, with API access and structured outputs for robotics and other physical-world workflows. It also supports image reasoning tasks such as pointing, counting, OCR, and document extraction.
MiniMax M3 is a frontier coding and agentic model with native multimodal understanding and up to 1M tokens of context. It is built for long-running developer workflows and is available through MiniMax’s API, MiniMax Code, and Token Plan.
Edgee Fallback Models réoriente automatiquement vos sessions Claude Code vers un autre modèle en cas de panne Anthropic ou de quota atteint, sans changer de code.
SemanticGuard is an AI gateway with a self-validating cache for LLM APIs. It helps developers and teams measure savings, reduce repeat request costs, and monitor cache correctness across providers such as OpenAI, Anthropic, and Google.
Gello is an Android app that runs a Hugging Face LLM locally and exposes it as a Discord bot. It lets people talk to the model in-channel on a phone-hosted setup without using a cloud API key.
TrackNotch is a native macOS app that tracks LLM usage across providers such as Claude, OpenAI, Cursor, Codex, Anthropic API, and Google Gemini. It displays that status in the notch or menu bar and keeps monitoring local to the machine.
Recall is a Chrome extension that tracks Claude.ai context usage, quota limits, and truncation risk in real time. It helps Claude users see when to summarize, split, or restart a conversation before responses degrade.
PromptQuorum is a browser-based tool for optimizing prompts, dispatching one prompt to 25+ AI models, and comparing the results with consensus and hallucination analysis. It supports cloud providers, local LLMs, and user-owned API keys.
Franz is a functional, prototype-oriented programming language hosted on GitHub. It is interpreted and dynamically typed, with documented LLVM native compilation, scoped closures, and a standard library for common language tasks.
Gemini 3.1 Flash-Lite is a Google Cloud Gemini model for low-latency, high-volume enterprise workflows on the Gemini Enterprise Agent Platform. It is positioned for agentic tasks, automated pipelines, and production use cases where speed and cost control matter.