Models & LLMs
Frontier models and LLMs I reach for in different workflows.
Gemini 3.0 Pro – Multimodal & Experiments
The latest Gemini Pro model I use for multimodal experiments, long-context reasoning, and building custom Gems and workflows.
Claude Sonnet 4.5 – Agentic Coding & Tools
Anthropic’s latest Sonnet model, described as the best coding model in the world and strongest for building complex agents, with major gains in reasoning, math, and real-world computer use while maintaining strong alignment. See the official announcement for details.
GPT‑5.2 – Generalist Frontier LLM
Latest GPT series frontier model from OpenAI, used as a default for mixed reasoning, coding, and product work across text and multimodal tasks.
Kimi K2.5 – Visual Agentic Intelligence
Open-source visual agentic intelligence with SOTA results on agentic, vision, and coding benchmarks. Includes Agent Swarm (beta) for up to 100 parallel sub-agents and 1,500 tool calls for 4.5× faster workflows. Useful links: - API: https://platform.moonshot.ai - Tech blog: https://kimi.com/blogs/kimi-k2-5.html - Weights & code: https://huggingface.co/moonshotai/Kimi-K2.5/tree/main