Working with models
How to choose a model
You always choose the model — in the model field. The service picks nothing for you. Below is a step-by-step guide: from the id format to matching a model to your task.
Step 1. The id format
A model id has the form provider/name, e.g. anthropic/claude-sonnet-4.6 or openai/gpt-4o. Case and exact spelling matter — copy the id verbatim.
Step 2. Fetch the model list
The live list with prices is always available:
curl https://api.alicode.store/v1/models -H "Authorization: Bearer sk-ali-•••"
The same list is available visually in the dashboard “Models” tab — search, vendor filters, price, vision badge, context size and a copy-id button. Models that are known to be non-functional are excluded.
Step 3. Match a model to your task
Rough guidance (see /v1/models for the live set):
| Model id | Provider | Best for |
|---|---|---|
| anthropic/claude-sonnet-4.6 | Anthropic | Daily coding, agents, tool use |
| anthropic/claude-3.5-haiku | Anthropic | Fast cheap loops, classification |
| openai/gpt-4o | OpenAI | All-rounder, vision, reliable JSON |
| openai/gpt-4o-mini | OpenAI | Cheapest OpenAI for volume |
| google/gemini-2.5-pro | Huge context, multimodal | |
| deepseek/deepseek-v3.1 | DeepSeek | Strong coding for pennies |
Selection criteria:
- Cost — input/output rate is shown in /v1/models;
- Context — max input the model accepts;
- Vision — multimodal models only (badge in the catalog);
- Task — Claude/DeepSeek for code/agents, mini models for cheap volume.
Put the chosen id into the model — that exact model is called.