Every big model. One key.

List prices, read live. Your allowance is in dollars, so a cheaper model stretches it further.

Claude Haiku 5.5
anthropic/claude-haiku-5.5
3,333 msgs / $1
$0.10 in · $0.50 out
Claude Sonnet 5.5
anthropic/claude-sonnet-5.5
166 msgs / $1
$2.00 in · $10.00 out
Claude Opus 5.5
anthropic/claude-opus-5.5
83 msgs / $1
$4.00 in · $20.00 out
Claude Fable 5.2
anthropic/claude-fable-5.2
41 msgs / $1
$8.00 in · $40.00 out
GPT-6 Luna
openai/gpt-6-luna
3,333 msgs / $1
$0.10 in · $0.50 out
GPT-6.1 Terra
openai/gpt-6.1-terra
333 msgs / $1
$1.00 in · $5.00 out
GPT-6.1 Sol
openai/gpt-6.1-sol
166 msgs / $1
$2.00 in · $10.00 out
Gemini 3.8 Flash
google/gemini-3.8-flash
404 msgs / $1
$0.83 in · $4.13 out
Gemini 4 Hydrogen
google/gemini-4-hydrogen
222 msgs / $1
$1.50 in · $7.50 out
Grok 4.7
x-ai/grok-4.7
250 msgs / $1
$2.00 in · $6.00 out
DeepSeek V4.1 Flash
deepseek/deepseek-v4.1-flash
1,333 msgs / $1
$0.30 in · $1.20 out

Text in, text out for now. A message is counted as 500 tokens in and 500 out.

Change one line. Keep your stack.

If your code already talks to OpenAI, point it at TOKENS and swap the key. Streaming, tool calls and JSON mode pass straight through.

POST /v1/chat/completions
Chat, with streaming
GET /v1/models
What is on, and what it costs
402
Today’s allowance is used
403
The wallet dropped under the minimum
Read the docs
cURL
Python
Node
Editors
curl https://www.holdtokens.fun/v1/chat/completions \
  -H "Authorization: Bearer $TOKENS_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-haiku-5.5",
    "messages": [{"role": "user", "content": "gm"}]
  }'