DeepSeek
DeepSeek offers some of the cheapest and most capable models for coding. Strong performance at a fraction of the cost of frontier providers.
Setup
# 1. Get your API key from https://platform.deepseek.com/
export DEEPSEEK_API_KEY=sk-...
# 2. Verify detection
arcana doctor
# 3. Run
arcana run "hello"
Available Models
| Model | Context | Best For |
|---|
deepseek-chat | 128K | General purpose, coding, very cheap |
deepseek-reasoner | 128K | Deep reasoning, math, complex analysis |
Config File
{
"provider": "deepseek",
"model": "deepseek-chat"
}
Environment Variables
| Variable | Description |
|---|
DEEPSEEK_API_KEY | Your DeepSeek API key (required) |
Pricing Notes
DeepSeek is extremely affordable. Approximate costs per 1M tokens:
| Model | Input | Output |
|---|
| deepseek-chat | $0.14 | $0.28 |
| deepseek-reasoner | $0.55 | $2.19 |
Tips
deepseek-chat is excellent for everyday coding at extremely low cost- Use
deepseek-reasoner for complex reasoning tasks — much cheaper than o1 or Claude Opus
- DeepSeek models are OpenAI-compatible — they work with OpenAI client libraries
- Great choice as a utility model for extraction and compaction tasks
Last updated: Jul 24, 2026
ARCANA