Providers
Model providers available through Rawrter.
Antropic
Anthropic’s flagship line is Claude Opus, designed for demanding analysis, system architecture, software engineering, and agentic tasks where reasoning quality is paramount. Claude Sonnet provides a strong quality-to-speed balance, while Haiku targets high-throughput requests. For long research, large codebases, and document collections, Claude Sonnet offers a context mode of up to 1 million tokens. This makes the Claude family especially useful for multi-step tool use, code review, knowledge synthesis, and precise business writing.
DeepSeek
DeepSeek is a Chinese AI company developing some of the world's most efficient AI models. It releases open-weight models under the MIT license that match the performance levels of leading global laboratories. DeepSeek was founded in 2023 by the Chinese hedge fund High-Flyer. The company quickly rose to the ranks of the world's top AI labs by releasing a series of models comparable in quality to GPT-4 and Claude, yet achieved at significantly lower training costs.
Gemini 3.1 Pro is Google’s flagship model for deep reasoning, complex problem solving, and reliable agentic workflows. Its key advantage is a context window of up to 1,048,576 input tokens and 65,536 output tokens, allowing a single request to retain large document sets, repositories, images, audio, video, and PDFs. The Gemini Flash line targets faster, more scalable workloads. Choose Gemini Pro for long research, large-codebase analysis, multimodal understanding, planning, and tool-using agents.
MiniMax
MiniMax M3 is MiniMax’s flagship model for software engineering and long-running agentic workflows. It combines strong coding and reasoning capabilities with native multimodality and a context window of up to 1 million tokens; the API guarantees at least 512,000 tokens. This scale allows a single conversation to retain large repositories, long documents, logs, images, and video. MiniMax M3 is best suited to autonomous agent tasks, software development and optimisation, deep material analysis, and scenarios where preserving context over a long sequence of steps is critical.
Moonshot AI (Kimi)
Moonshot AI is a Chinese AI startup and the creator of the Kimi model series. It specializes in models featuring long-context capabilities and advanced agentic functions. Founded in Beijing in 2023, Moonshot AI focuses on models with long-context capabilities and advanced agentic functions. The company ranks among the top 10 global AI startups in terms of model quality. Kimi K2.5 (released January 27, 2026) is a flagship multimodal model featuring 1 trillion parameters (32B active) and 384 experts, trained on 15 trillion mixed visual and text tokens. It achieves scores of 85.0% on LiveCodeBench (surpassing Claude Opus 4.5 at 64.0%), 96.1% on AIME 2025, 95.4% on HMMT, and 92.3% on OCRBench. It utilizes Agent Swarm technology, supporting up to 100 parallel sub-agents and 1,500 simultaneous tool calls.
OpenAI
OpenAI’s flagship model is GPT‑5.6 Sol, designed for demanding work across coding, knowledge work, research, science, and agentic workflows. GPT‑5.6 Sol Pro is available for the most difficult and long-running tasks; GPT‑5.6 Terra provides a balanced capability, speed, and cost profile, while GPT‑5.6 Luna targets minimal latency and cost-efficient high-volume workloads. The ecosystem’s strength is the ability to choose reasoning effort and a model for the task, while the available context window depends on the selected model and API. Choose OpenAI for code generation and review, analysis of large documents and data, multi-step planning, assistants, multimodal interfaces, and workflow automation.
Qwen (Alibaba)
Qwen is a series of powerful AI models from Alibaba Cloud. It offers world-class performance, open weights, and support for a wide range of tasks, from text processing to programming.
Xiaomi
MiMo‑V2.5‑Pro is Xiaomi’s flagship reasoning model for complex tasks, software engineering, and reliable agents. It supports up to 1 million context tokens and up to 128,000 output tokens, so a single request can retain substantial documents, repositories, and long work histories. MiMo‑V2.5 complements it with native text, image, video, and audio understanding at the same million-token context scale. The MiMo family is especially effective for technical research, code generation and debugging, multi-step planning, large material collections, and agents that must retain long context.
Z.ai
GLM‑5 is Z.ai’s flagship line for agentic engineering, programming, and complex reasoning. GLM‑5 supports a 200,000-token context window and up to 128,000 output tokens, enough for large repositories, technical documentation, and long sequences of tool calls. Faster variants in the line can reduce latency in interactive workflows. Choose GLM for code generation and refactoring, debugging, engineering analysis, multi-step agents, and process automation where reliably retaining a large body of context matters.