The smallest model in DeepSeek's new architecture family, with native image understanding. It runs in thinking mode by default and also offers a non-thinking mode; DeepSeek reports it outperforms its earlier DeepSeek-V4-Pro flagship on benchmarks.
per 1M tokens
Latency3.7sSpeed103 tok/s
The first natively multimodal model in Z.ai's GLM-5 series, taking text, image, video and file input. Z.ai positions it as stronger than GLM-5.2 at a much lower cost.
per 1M tokens
Latency7.1sSpeed38 tok/s
Z.ai's flagship GLM model for complex software engineering and agentic work. It keeps GLM-5.2's base model, with post-training gains in coding, long-horizon tasks and cybersecurity, and currently accepts text input only.
per 1M tokens
Latency9.2sSpeed36 tok/s
Moonshot AI's flagship open-weight model, with 2.8 trillion parameters built on Kimi Delta Attention and Attention Residuals. It has native visual understanding and is designed for long-horizon coding, knowledge work and deep reasoning.
per 1M tokens
Latency9.5sSpeed30 tok/s
xAI's most capable model, for coding, agentic tasks and knowledge work, released in September 2026 as the successor to Grok 4.6. It takes text and image input with a 500K-token context window and offers four reasoning effort levels, from low to xhigh.
per 1M tokens
Latency7.8sSpeed59 tok/s