CoralBricks

Overview

Property	Details
Description	CoralBricks serves open-weight GLM and DeepSeek models in FP4 over an OpenAI-compatible API
Provider Route on LiteLLM	coralbricks/
Link to Provider Doc	CoralBricks
Default Base URL	https://inference.coralbricks.ai/v1
Supported Operations	/chat/completions, /responses, /messages

CoralBricks is its own provider on LiteLLM rather than a generic OpenAI-compatible route, so its spend is priced from the CoralBricks cost map entries and reported under coralbricks instead of being pooled with OpenAI traffic. /responses and /messages requests are forwarded to the matching CoralBricks endpoints as they arrive, with no translation through chat completions

API Key

Environment Variables

import os

os.environ["CORALBRICKS_API_KEY"] = "your-api-key"
os.environ["CORALBRICKS_API_BASE"] = "https://inference.coralbricks.ai/v1"  # optional override

Models

Model	Input / 1M tokens	Output / 1M tokens	Cache write / 1M tokens	Cache read / 1M tokens
coralbricks/glm-5.3-fp4	$1.12	$4.40	$1.68	$0
coralbricks/glm-5.3-flash-fp4	$0.15	$0.50	$0.23	$0
coralbricks/deepseek-v4.1-flash-fast-fp4	$0.30	$1.20	$0.09	$0

Every model takes up to 1,048,576 input tokens and supports tool calling, reasoning, and prompt caching. Pricing follows the CoralBricks pricing page: cache writes are billed at the cache write rate and cached reads are free. If your contract prices differ, set input_cost_per_token / output_cost_per_token on the deployment and those override the cost map

Usage - LiteLLM Python SDK

Chat Completions

CoralBricks Chat Completion

import os
from litellm import completion

os.environ["CORALBRICKS_API_KEY"] = "your-api-key"

response = completion(
    model="coralbricks/glm-5.3-flash-fp4",
    messages=[{"role": "user", "content": "Write a python function that reverses a string"}],
)

print(response.choices[0].message.content)

Streaming

CoralBricks Streaming Chat Completion

import os
from litellm import completion

os.environ["CORALBRICKS_API_KEY"] = "your-api-key"

response = completion(
    model="coralbricks/glm-5.3-flash-fp4",
    messages=[{"role": "user", "content": "Explain a binary search in two sentences"}],
    stream=True,
)

for chunk in response:
    print(chunk)

Responses API

CoralBricks Responses API

import os
import litellm

os.environ["CORALBRICKS_API_KEY"] = "your-api-key"

response = litellm.responses(
    model="coralbricks/glm-5.3-flash-fp4",
    input="Explain a binary search in two sentences",
    max_output_tokens=256,
)

print(response.output_text)

Usage - LiteLLM Proxy

Add CoralBricks to your LiteLLM Proxy configuration:

config.yaml

model_list:
  - model_name: glm-5.3-flash
    litellm_params:
      model: coralbricks/glm-5.3-flash-fp4
      api_key: os.environ/CORALBRICKS_API_KEY

general_settings:
  master_key: os.environ/LITELLM_MASTER_KEY

Start the proxy:

Start LiteLLM Proxy

export CORALBRICKS_API_KEY="your-api-key"
export LITELLM_MASTER_KEY="sk-local-coralbricks"
litellm --config config.yaml --port 4000

# RUNNING on http://0.0.0.0:4000

OpenAI SDK

CoralBricks via Proxy - OpenAI SDK

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:4000",
    api_key="sk-local-coralbricks",
)

response = client.chat.completions.create(
    model="glm-5.3-flash",
    messages=[{"role": "user", "content": "hello from litellm"}],
)

print(response.choices[0].message.content)

cURL

CoralBricks via Proxy - cURL

curl http://localhost:4000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $LITELLM_MASTER_KEY" \
  -d '{
    "model": "glm-5.3-flash",
    "messages": [{"role": "user", "content": "hello from litellm"}]
  }'

You can also add CoralBricks from the Admin UI. Go to Models, then Add Model, pick CoralBricks as the provider, choose one of the coralbricks/ models, and paste your key

Responses API

The proxy forwards /v1/responses requests to the CoralBricks Responses endpoint:

Responses API through LiteLLM Proxy

curl http://localhost:4000/v1/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $LITELLM_MASTER_KEY" \
  -d '{
    "model": "glm-5.3-flash",
    "input": "hello from litellm",
    "max_output_tokens": 128
  }'

Anthropic Messages Compatibility

The proxy forwards /v1/messages requests to the CoralBricks Messages endpoint, so Anthropic SDK clients work unchanged:

Anthropic Messages through LiteLLM Proxy

curl http://localhost:4000/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: $LITELLM_MASTER_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -d '{
    "model": "glm-5.3-flash",
    "max_tokens": 128,
    "messages": [{"role": "user", "content": "hello from litellm"}]
  }'

Cost Tracking

The coralbricks/ models are registered in LiteLLM's model cost map, so per-request spend is computed automatically on all three endpoints, returned in the x-litellm-response-cost response header, and recorded in spend logs under provider coralbricks. Cache writes and cached reads are tracked from the usage CoralBricks returns, so a repeated prompt is billed at the free cached read rate for its cached prefix

Custom Endpoints

Set CORALBRICKS_API_BASE or pass api_base explicitly to route through a different CoralBricks endpoint. The coralbricks/ route keeps the provider identity and pricing either way

config.yaml

model_list:
  - model_name: glm-5.3-flash
    litellm_params:
      model: coralbricks/glm-5.3-flash-fp4
      api_base: https://your-coralbricks-endpoint/v1
      api_key: os.environ/CORALBRICKS_API_KEY