apillms vs OpenRouter
Same models, one key, lower price. USD per 1M tokens. apillms prices are live; OpenRouter prices are their public list prices (snapshot 29 Sep 2026).
| Model | apillms in | OpenRouter in | apillms out | OpenRouter out | apillms cache | OpenRouter cache | You save* |
|---|---|---|---|---|---|---|---|
| Loading… | |||||||
* Savings for a typical request of 3 input tokens per 1 output token, without cache.
Save tokens: how to pay less
1. Put stable text first
System prompt, tool definitions and long documents go at the start and stay byte-for-byte identical. apillms adds a per-user cache key automatically, so repeated prefixes are billed at the cache price.
2. Use Claude for long contexts
Claude models give the most reliable cache hits (often 70%+ of input on repeat turns). Claude Code and Cursor send cache markers for you.
3. Cap max_tokens
You are only billed for output up to your max_tokens. Set it to what you actually need.
4. Pick the right tier
Use Haiku, GPT Luna or DeepSeek Flash for routing, extraction and summaries; keep Opus / GPT-6 for hard reasoning.
FAQ
Is apillms cheaper than OpenRouter?
For most models, yes. The table on this page compares live apillms prices with OpenRouter's public list prices per million input, output and cached tokens.
How does prompt caching save money?
When the start of your prompt (system prompt, tools, long documents) repeats, cached tokens are billed at a fraction of the normal input price. Keep stable content first and changing content last.
Do I need a subscription?
No. Buy credits from $5 with PayPal or card; credits never expire.