The model “Kimi K2 Thinking” is a new reasoning-agent variant developed by Moonshot AI (Beijing). It belongs to the broader “Kimi K2” family of large-language models but is specifically tuned for thinking—i.e., long-horizon reasoning, tool usage, planning and multi-step inference. Version are kimi-k2-thinking-turbo，kimi-k2-thinking. Basic Features Technical Details Architecture: Training / optimisation: Inference & tool-use: Benchmark […]
Model Type: Chat
Kling 2.5 Turbo released late September 2025 — and positioned as a fast, cinematic text-to-video / image-to-video model. The release emphasizes faster generation, improved prompt adherence, and better temporal / camera control, positioning the model as a production-oriented upgrade over earlier Kling 2.x releases.
Model Type: Video
MiniMax M2 is an open-source, agent-native large language model (LLM) released by MiniMax on October 27, 2025. It is explicitly engineered for coding and agentic workflows (tool calling, multi-step automation), prioritizing low latency and cost-effective serving while delivering strong reasoning and tool-use capabilities.
Model Type: Chat
GLM-4.6 is the latest major release in Z.ai’s (formerly Zhipu AI) GLM family: a 4th-generation, large-language MoE (Mixture-of-Experts) model tuned for agentic workflows, long-context reasoning and real-world coding. The release emphasizes practical agent/tool integration, a very large context window, and open-weight availability for local deployment.
Model Type: Chat
Claude Haiku 4.5 is a purpose-optimized, smaller-class language model from Anthropic, released in mid-October 2025. It’s positioned as a fast, low-cost option in the Claude lineup that preserves strong capability on tasks like coding, agent orchestration, and interactive “computer-use” workflows while enabling much higher throughput and lower unit cost for enterprise deployments.
Model Type: Chat
Sora 2 is OpenAI’s flagship text-to-video and audio generation system designed to produce short cinematic clips with synchronized dialogue, sound effects, persistent scene state, and markedly improved physical realism. Sora 2 represents OpenAI’s step forward in producing short, controllable videos with synchronized audio (speech and sound effects), improved physical plausibility (motion, momentum, buoyancy), and stronger safety controls compared with earlier text-to-video systems.
Model Type: Video
Veo 3.1 is Google’s incremental-but-significant update to its Veo text-and-image→video family, adding richer native audio, longer and more controllable video outputs, and finer editing and scene-level controls.
Model Type: Video
GPT-5-Codex is a specialized variant of OpenAI’s GPT-5 family designed for complex software engineering workflows: coding, large-scale refactoring, long multi-step agentic tasks, and extended autonomous runs inside the Codex environment (CLI, IDE extension, and cloud). It is positioned as the default model for OpenAI’s Codex product and is accessible via the Responses API and Codex subscriptions.
Model Type: Code
Sora 2 Pro is OpenAI’s flagship video+audio generation model designed to create short, highly realistic video clips with synchronized dialogue, sound effects, and stronger physical/world simulation than previous video models. It’s positioned as the higher-quality “Pro” variant available to paying users and via the API for programmatic generation. The model emphasizes controllability, temporal coherence, and audio synchronization for cinematic and social use cases.
Model Type: Video
GPT-5 Pro is OpenAI’s top-tier GPT-5 variant designed for extended reasoning, code-quality improvements, and higher-stakes workflows. It sits at the apex of the GPT-5 family (which also includes standard and “Thinking” modes) and is positioned for developers and organizations that need the strongest single-query reasoning and task performance available in the GPT-5 lineup.
Model Type: Chat
