One API Key.
91 LLMs.

Yielding Bear exposes 91 models on one OpenAI-compatible endpoint. Point your client at the base URL below — same request shape you already use.

With Max, Auto prefers a cheap rail for routine steps and only spends premium when the step needs it. Pin a model id anytime you want a fixed choice. Either way is the same endpoint.

Two modes: set model to a catalog id, or use yieldingbear/grizzly-max for Auto.

All API requests require an API key passed as a Bearer token in the Authorization header. Get your API key from the dashboard.

Authorization: Bearer yb_live_sk_...xxxxxxxxxxxxxxxxxxxxxxxx

export YIELDINGBEAR_API_KEY=  # YB_API_KEY also accepted
export YIELDINGBEAR_API_KEY="yb_live_sk_aaf1ccf8b34ebda1c52d..."

# Verify your key works
curl https://yieldingbear.com/api/v1/usage \
  -H "Authorization: Bearer $YIELDINGBEAR_API_KEY"

The Chat Completions endpoint is universally compatible. Send a POST request with messages and receive a structured response. Works with any LLM SDK, including the OpenAI SDKs.

POST /api/v1/chat/completions

{
  "model": "groq/llama-3.1-8b-instant",
  "messages": [
    { "role": "system", "content": "You are a helpful assistant." },
    { "role": "user", "content": "What is the capital of France?" }
  ],
  "temperature": 0.7,
  "max_tokens": 256,
  "stream": false
}

curl -X POST https://yieldingbear.com/api/v1/chat/completions \
  -H "Authorization: Bearer $YIELDINGBEAR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "groq/llama-3.1-8b-instant",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

{
  "id": "chatcmpl-3552055b-e963-4907-a077-d76f5a1019e7",
  "object": "chat.completion",
  "created": 1781641651,
  "model": "groq/llama-3.1-8b-instant",
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "The capital of France is Paris."
    },
    "finish_reason": "stop"
  }],
  "usage": {
    "prompt_tokens": 36,
    "completion_tokens": 24,
    "total_tokens": 36
  },
  "cost_usd": 0.000017,
  "balance_remaining_usd": 4.9911
}

{
  "id": "chatcmpl-3552055b-e963-4907-a077-d76f5a1019e7",
  "object": "chat.completion",
  "created": 1781641651,
  "model": "groq/llama-3-3-70b-versatile",   ← actual underlying model that served the request
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "It's nice to meet you. Is there something I can help you with, or would you like to chat?"
    },
    "finish_reason": "stop"
  }],
  "usage": {
    "prompt_tokens": 36,
    "completion_tokens": 24,
    "total_tokens": 36
  },
  "cost_usd": 0.000017,
  "balance_remaining_usd": 4.9911,
  "routing": {
    "category": "general",                                       ← which specialist was called
    "intent":   "text",                                          ← text or image
    "tier":     "low",                                           ← low = efficient, high = powerful
    "routed_to": "groq/llama-3-3-70b-versatile",                 ← resolved underlying model
    "reason":    "general:no-high-signals;default-low",          ← why this tier was picked
    "explicit":  false                                           ← true if user forced tier/force via header or body
  }
}

{
  "id": "chatcmpl-R1p2pOLmD2huJgdFQLrm0BCd",
  "model": "see live catalog",   ← swapped to the pin a $0 catalog model
  ...
  "routing": {
    "category": "coding",
    "intent":   "text",
    "tier":     "high",
    "routed_to": "see live catalog",
    "reason":    "cat=coding, kw:\"refactor\"",
    "explicit":  false
  }
}

Call the Grizzly Auto router (or pin any live catalog model id). The router classifies the prompt and picks the most efficient underlying model — paying nothing extra on top of the routed model's per-Mtok rate. Same endpoint, same auth, same response shape as Chat Completions. Returns an extra routing field so you can see what was picked.

curl https://yieldingbear.com/api/v1/chat/completions \
  -H "Authorization: Bearer $YIELDINGBEAR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "yieldingbear/grizzly-max",
    "messages": [
      { "role": "user", "content": "Write a haiku about coding" }
    ]
  }'

curl https://yieldingbear.com/api/v1/chat/completions \
  -H "Authorization: Bearer $YIELDINGBEAR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "yieldingbear/grizzly-max",
    "messages": [
      { "role": "user", "content": "Refactor this Python function to use async/await:\n\ndef fetch_all(urls):\n    return [requests.get(u).json() for u in urls]" }
    ]
  }'

curl https://yieldingbear.com/api/v1/chat/completions \
  -H "Authorization: Bearer $YIELDINGBEAR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "yieldingbear/grizzly-max",
    "messages": [
      { "role": "user", "content": "Write a 3-sentence product description for a bear-themed energy drink" }
    ]
  }'

curl https://yieldingbear.com/api/v1/chat/completions \
  -H "Authorization: Bearer $YIELDINGBEAR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "yieldingbear/grizzly-max",
    "messages": [
      { "role": "user", "content": "What is the current price of NVDA and what is its 50-day moving average?" }
    ]
  }'

Generate text embeddings through a single endpoint. Output is an OpenAI-shaped vector array you can drop into any vector DB (Pinecone, Supabase pgvector, Weaviate, etc.). Use the response data[].embedding field — no parsing needed.

POST /api/v1/embeddings

{
  "model": "deepinfra/bge-base-en-v1.5",
  "input": "The food was delicious and the service was excellent."
}

curl -X POST https://yieldingbear.com/api/v1/embeddings \
  -H "Authorization: Bearer $YIELDINGBEAR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepinfra/bge-base-en-v1.5",
    "input": "The food was delicious and the service was excellent."
  }'

{
  "object": "list",
  "data": [
    {
      "object": "embedding",
      "index": 0,
      "embedding": [0.0107, 0.0557, 0.0270, ...768 floats total]
    }
  ],
  "model": "deepinfra/bge-base-en-v1.5",
  "usage": { "prompt_tokens": 4, "total_tokens": 4 },
  "cost_usd": 0,
  "balance_remaining_usd": 4.99
}

List all available models. This endpoint is public and does not require authentication. Use it to discover models and check their current pricing.

GET /api/v1/models

curl https://yieldingbear.com/api/v1/models

Rate limits protect the API from abuse and ensure fair access. Limits are applied per API key and vary by tier.

{
  "error": {
    "message": "Rate limit exceeded: 60 requests per minute",
    "type": "rate_limit_exceeded",
    "code": "rate_limit_exceeded"
  }
}

Enable streaming to receive tokens as they are generated. This reduces perceived latency for long responses. Use the stream: true parameter.

{
  "model": "groq/llama-3.1-8b-instant",
  "messages": [{"role": "user", "content": "Write a haiku about coding"}],
  "stream": true
}

data: {"id":"chatcmpl-e96f1581-d9bb-46c9-8f9c-68b27e45a717","choices":[{"finish_reason":null,"index":0,"delta":{"function_call":null,"tool_calls":null,"content":"Lines","role":"assistant"}}],"created":1781641928,"model":"llama-3.1-8b-instant","object":"chat.completion.chunk","system_fingerprint":"fp_e2c608b1d6","usage":{}}

data: {"id":"chatcmpl-e96f1581-d9bb-46c9-8f9c-68b27e45a717","choices":[{"finish_reason":null,"index":0,"delta":{"function_call":null,"tool_calls":null,"content":" of"}}],"created":1781641928,"model":"llama-3.1-8b-instant","object":"chat.completion.chunk","system_fingerprint":"fp_e2c608b1d6","usage":{}}

data: {"id":"chatcmpl-e96f1581-d9bb-46c9-8f9c-68b27e45a717","choices":[{"finish_reason":"stop","index":0,"delta":{"content":null}}],"created":1781641928,"model":"llama-3.1-8b-instant","object":"chat.completion.chunk","system_fingerprint":null,"usage":{}}

data: [DONE]

import sys
import os
sys.path.insert(0, os.path.expanduser("~/.yieldingbear"))
from yieldingbear import YieldingBear

client = YieldingBear.from_key_file()  # picks up ~/.hermes/secrets/yieldingbear-token

for chunk in client.chat.completions.stream(
    model="nvidia/nemotron-3.5-lightning",   # or yieldingbear/grizzly-max Auto router
    messages=[{"role": "user", "content": "Hello!"}],
):
    if chunk.delta:                      # chunk.delta is a str of the new text
        print(chunk.delta, end="", flush=True)

Every chat response includes cost_usd and balance_remaining_usd. For an aggregated view (totals, per-model breakdown, recent requests), call GET /api/v1/usage with your API key.

GET /api/v1/usage

curl https://yieldingbear.com/api/v1/usage \
  -H "Authorization: Bearer $YIELDINGBEAR_API_KEY"

{
  "account": { "email": "you@example.com", "credit_balance_usd": 4.99 },
  "totals": {
    "cost_usd": 0.000452,
    "prompt_tokens": 1359,
    "completion_tokens": 1020,
    "total_tokens": 2379,
    "calls": 50
  },
  "by_model": [
    { "model": "yieldingbear/grizzly-max", "calls": 3, "tokens": 157, "cost_usd": 0.000031 },
    { "model": "anthropic/claude-3-5-sonnet-latest", "calls": 2, "tokens": 34, "cost_usd": 0.000306 }
  ],
  "usage": [
    {
      "id": "edfcb790-d59e-4f51-b543-f45bf39bb3e9",
      "model": "groq/llama-3.1-8b-instant",
      "provider": "groq",
      "input_tokens": 36,
      "output_tokens": 8,
      "total_tokens": 44,
      "cost_usd": 0,
      "latency_ms": 137,
      "created_at": "2026-06-16T20:27:16.396812+00:00"
    }
  ]
}

The API follows OpenAI's error format. All errors return a JSON body with a top-level error object. Always check for errors in your response handling.

{
  "error": {
    "message": "Invalid API key",
    "type": "invalid_request_error",
    "code": "authentication_error",
    "param": null,
    "status": 401
  }
}

Free: unlimited calls on the low-tier free model (nvidia/nemotron-3.5-lightning). Credits or Max open the rest of the catalog (OpenAI, Anthropic, Google, DeepSeek, Mistral, Groq, xAI, …) at a small markup over provider rates.

Start with the native Yielding Bear SDK — zero dependencies, vendored by the installer. Every other LLM SDK works too: just point its base_url at https://yieldingbear.com/api/v1.

Yielding Bear SDK

Zero dependencies. Vendored to ~/.yieldingbear/lib/ by the installer.

Python

Node