If you use a router to reach many models with one key, Tokenless does the same with frontier Claude and GPT, but bills every token at 50% of the provider's list price instead of passing list price through with a markup.
| Model | Input / 1M | Output / 1M | Context |
|---|---|---|---|
| Claude Fable 5Anthropic | $5.00 $10.00 | $25.00 $50.00 | 1,000,000 |
| GPT-5.6 SolOpenAI | $2.50 $5.00 | $15.00 $30.00 | 1,000,000 |
| GPT-5.6 TerraOpenAI | $1.25 $2.50 | $7.50 $15.00 | 400,000 |
| GPT-5.6 LunaOpenAI | $0.50 $1.00 | $3.00 $6.00 | 400,000 |
| Claude Sonnet 5Anthropic | $1.50 $3.00 | $7.50 $15.00 | 1,000,000 |
| GPT-5.5OpenAI | $2.50 $5.00 | $15.00 $30.00 | 1,000,000 |
| GPT-5.5 ProOpenAI | $15.00 $30.00 | $90.00 $180.00 | 1,000,000 |
| Claude Opus 4.8Anthropic | $2.50 $5.00 | $12.50 $25.00 | 1,000,000 |
| GPT-5.4OpenAI | $1.25 $2.50 | $7.50 $15.00 | 400,000 |
| GPT-5.4 ProOpenAI | $15.00 $30.00 | $90.00 $180.00 | 400,000 |
| GPT-5.4 miniOpenAI | $0.38 $0.75 | $2.25 $4.50 | 400,000 |
| GPT-5.4 nanoOpenAI | $0.10 $0.20 | $0.63 $1.25 | 200,000 |
| Claude Opus 4.7Anthropic | $2.50 $5.00 | $12.50 $25.00 | 1,000,000 |
| GPT-5.3 CodexOpenAI | $0.88 $1.75 | $7.00 $14.00 | 400,000 |
| Claude Opus 4.6Anthropic | $2.50 $5.00 | $12.50 $25.00 | 1,000,000 |
| Claude Sonnet 4.6Anthropic | $1.50 $3.00 | $7.50 $15.00 | 1,000,000 |
| Claude Haiku 4.5Anthropic | $0.50 $1.00 | $2.50 $5.00 | 200,000 |
| Gemini 3.1 ProGoogle | $1.00 $2.00 | $6.00 $12.00 | 1,000,000 |
| Gemini 3.5 FlashGoogle | $0.25 $0.50 | $1.75 $3.50 | 1,000,000 |
| Gemini 3.1 Flash-LiteGoogle | $0.08 $0.15 | $0.30 $0.60 | 1,000,000 |
| Gemini 2.5 ProGoogle | $0.63 $1.25 | $5.00 $10.00 | 1,000,000 |
| Gemini 2.5 FlashGoogle | $0.15 $0.30 | $1.25 $2.50 | 1,000,000 |
| Gemini 2.5 Flash-LiteGoogle | $0.05 $0.10 | $0.20 $0.40 | 1,000,000 |
Prices are USD per million tokens. Struck-through figures are the provider's official list price; the bold figure is the Tokenless rate at 50% of list.
Tokenless is OpenAI-compatible. Keep your existing SDK and code, swap the base URL, and every request bills at half of list price. The same key reaches every Claude and GPT model.
from openai import OpenAI
client = OpenAI(
api_key="YOUR_TOKENLESS_KEY",
base_url="https://api.tokenless.store/api/v1", # the only line you change
)
resp = client.chat.completions.create(
model="opus-4.8",
messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)Both give you one key and one OpenAI-compatible endpoint for many models. The difference is price: Tokenless charges 50% of the provider's list price per token rather than passing list price through with an added margin.
The frontier Claude and GPT families, including Claude Fable 5, Opus 4.8, and Sonnet, plus the GPT-5.6, GPT-5.5, and GPT-5.4 lines. See the full list on the models page.
No. Tokenless is OpenAI-compatible. Point your existing SDK at the Tokenless base URL and keep your requests as they are.
No. Tokenless does not log prompts or completions and does not use your data for training. Keys are hashed at rest.
Yes. You prepay a balance and draw it down per token. Requests stop at zero balance, so there is no debt. Your first $1 is free.
Create a key, add the base URL, and your first $1 of usage is free. No subscription, no minimums.