More choices have landed in the Pickaxe model selector! Whether you’re building a quick-response assistant or an agent that works through longer, tool-driven tasks, there’s something here worth testing against your existing prompts!
OpenAI
- GPT-6.1 Sol is an updated option for complex coding and professional work. Its long context and configurable reasoning make it worth testing on agents that need to work through substantial reference material or solve a problem in several steps.
- GPT-6 Sol is designed for demanding coding and agentic workflows. Try it when an agent needs to reason through a request, use Actions, and bring the results together in a useful response.
- GPT-6 Luna is the efficient choice for focused, high-volume tasks. It could suit assistants handling frequent questions, drafting short responses, or sorting and extracting information where speed and cost matter.
Anthropic
- Claude Haiku 5.5 is a low-latency option for classification, extraction, and routing. It supports adaptive reasoning, so it’s worth trying for workflows that process many requests but still occasionally need a more considered answer.
- Claude Sonnet 5.5 balances speed and capability for coding, Actions, and everyday business work. It’s a candidate for agents that need to read lengthy material, follow detailed instructions, and take the next step with a tool.
xAI
- Grok 4.7 is aimed at coding, agentic tasks, and knowledge work. With configurable reasoning and text and image input, it’s another model to compare on workflows that combine visual context with multi-step questions.
Xiaomi MiMo
- MiMo V2.6 Flash is the efficient member of the family, built for reasoning, coding, and high-volume agent workflows. It’s worth testing where an agent runs often and you want to keep a close eye on cost per response.
- MiMo V2.6 Pro is geared toward complex coding, research, and longer tool-driven workflows. Try it on requests that require the agent to gather information through Actions and work across multiple steps before answering.
- MiMo V2.6 Pro UltraSpeed uses the Pro model checkpoint with a higher-throughput option. It gives builders a way to test the same Pro-style workflows when getting responses out faster is a priority.
We’re excited to see what these new models can do. All nine of these models are available in the builder! Please try them out and let us know how you’re utilizing them to improve the performance of your agents!