You can use frontier reasoning models, code-specialized models, and lower-cost models for agentic coding, depending on what your tool supports. The best model depends on the task: architecture needs reasoning, implementation needs code quality, and routine cleanup often needs cost efficiency.
Verdent's public pricing and model-pricing pages list models across Anthropic, OpenAI, Google, Moonshot, Zhipu AI, and MiniMax, but exact model names can differ between pricing surfaces and should be rechecked before publishing.
For BYOK, Verdent documents Anthropic, OpenAI, and OpenRouter as supported providers. That matters because model choice should not be locked to one vendor. Use stronger models for difficult code reasoning, cheaper models for routine work, and always recheck the current official model list before publishing exact names.
