Mistral-Next

« An open-source model designed for code and long-running tasks, with a context window of up to 1 million tokens. This Mixed Expert (MoE) LLM handles complex projects, analyzes entire repositories, and performs multi-step reasoning. It is available as open source under the MIT license »

« Discover a next-generation AI model designed for advanced use while remaining highly secure. You’ll enjoy top-tier performance for coding, analysis, and research, while safeguards block sensitive applications such as cybersecurity and biology »

« This Mixture-of-Experts model, with approximately 1 trillion parameters, features a 1M-token context, virtually infinite Engram memory, and multimodal capabilities for text, images, and video. It aims to achieve performance scores comparable to Claude Opus while remaining significantly more affordable (to be released under the Apache 2.0 license) »

« Designed for intensive tasks, this model from Google’s Gemini family outperforms its predecessor in token efficiency while improving performance in coding and knowledge-based tasks. With a 1-million-token context, full multimodality, and better agentic task performance »

« Take advantage of Anthropic’s AI model—more reliable, more honest, and four times less likely to overlook its own errors. Its fast mode delivers up to 2.5 times more tokens per second (at the same API rate as the previous version) »

« Moonshot's LLM model, which rivals the best proprietary models on numerous benchmarks. Designed for long-form coding, advanced reasoning, and knowledge work tasks, it is also the first open-source model to surpass the 2.8 trillion-parameter mark »

« Tackle time-consuming and complex tasks by letting AI plan, use tools, check its work, and see tasks through to completion with minimal supervision. This model significantly improves coding, the use of virtual machines, online research, and knowledge work tasks »

« Anthropic's latest high-end model, optimized for agentic code, long-running tasks, and critical business scenarios. It features self-checking of responses and a context window of one million tokens. It is also capable of automatically adjusting the depth of its reasoning based on the difficulty of the problem »