API
Models
Every id we accept and how it maps to price tiers.
Models
Token Harbor accepts any model id from OpenAI, Anthropic, Google, DeepSeek, and a curated OpenRouter list. You don't sign up with each provider separately — one Universal Key, one balance, one list.
Pricing tiers
Every model maps to one of three tiers. Price is per completed reply, not per token — no surprise bills from verbose models.
| Tier | Web | API | Cashback on paid balance |
|---|---|---|---|
| Low | $0.01 | $0.03 | 90% |
| Mid | $0.05 | $0.15 | 70% |
| High | $0.15 | $0.45 | 50% |
See Cashback for the full formula.
Popular ids
| Id | Tier | Provider |
|---|---|---|
gpt-5 | High | OpenAI |
gpt-4.1 | Mid | OpenAI |
gpt-4o | Mid | OpenAI |
gpt-4o-mini | Low | OpenAI |
claude-opus-4-5 | High | Anthropic |
claude-sonnet-4-5 | Mid | Anthropic |
claude-haiku-4-5 | Low | Anthropic |
deepseek-chat | Low | DeepSeek |
deepseek-reasoner | Mid | DeepSeek (R1) |
gemini-2.5-pro | High | |
gemini-2.5-flash | Low |
Anything else we don't explicitly list is accepted as-is and routed to OpenRouter's catalog — we do our best to match it to a tier, but if in doubt we default to Mid.
Aliases
Some ids have historical aliases that still work:
| Old | New |
|---|---|
gpt-4 | gpt-4o |
gpt-3.5-turbo | gpt-4o-mini |
claude-3-opus | claude-opus-4-5 |
claude-3-sonnet | claude-sonnet-4-5 |
Context limits
We never truncate your conversation. Each request goes straight to
the upstream; if the upstream refuses (context_length_exceeded),
you'll see that error verbatim.