Ring-2.6-1T

Ring-2.6-1T is InclusionAI's trillion-parameter reasoning model built for real-world agent workflows, using a sparse Mixture-of-Experts architecture with roughly 63 billion active parameters per token and a 262K context window. Released in May 2026 under the MIT license, this open-weight model adds an adjustable reasoning-effort mechanism with high and xhigh modes, letting developers trade thinking depth against token cost by task complexity. Ring-2.6-1T reports leading results on reasoning and agentic benchmarks: 95.83 on AIME 2026, 88.27 on GPQA Diamond, and 87.60 on PinchBench, where InclusionAI notes it edges past GPT-5.4 and Gemini 3.1 Pro, alongside 95.32 on the TAU2-Bench Telecom scenario. The model focuses on multi-step task planning, tool invocation, error correction, and continuous execution rather than isolated question answering, which suits coding agents and complex automation. High mode favors frequent agent interactions and tool collaboration with reduced reasoning overhead, while xhigh mode opens extensive reasoning space for mathematics, scientific analysis, and multi-path exploration. As a thinking model, Ring-2.6-1T produces reasoning output and is available on Hugging Face with support for SGLang and vLLM.