Docs · v1

Five minutes to first token.

OpenAI-compatible and Anthropic-compatible endpoints, one key, prepaid credits, seven curated uncensored builds. If your tool can change a base URL, it works here.

01 Quickstart

Grab a key in the console, top up $5, then:

# curl
curl https://api.cheapweights.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-…" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3.8-27b",
    "messages": [{"role":"user","content":"hello"}],
    "stream": true
  }'
# python (openai SDK, unchanged)
from openai import OpenAI
client = OpenAI(base_url="https://api.cheapweights.ai/v1", api_key="sk-…")
print(client.chat.completions.create(model="gpt-oss-120b",
      messages=[{"role":"user","content":"hello"}]).choices[0].message.content)

02 Model IDs

Model IDs match the price sheet. Every model is an uncensored build, so ids are bare:

Older docs and scripts that append :unc or :uncensored keep working: the gateway aliases both to the bare id. Unknown model ids return 404; we never silently substitute a different model or a different price.

03 Endpoints

EndpointShapeNotes
POST /v1/chat/completionsOpenAIstreaming, tool calling, JSON mode on all self-hosted models
GET /v1/modelsOpenAIlive model ids; prices live in /catalog.json and on the price sheet
POST /v1/messagesAnthropicunified endpoint — any model on the sheet; forwards anthropic-version/anthropic-beta headers
GET /v1/account/balanceCheapWeightscredits remaining for the account behind the key: { credits, usd, granted_usd, spent_usd }
POST /v1/embeddingsOpenAIweek 2, with the embeddings models

04 Agent harnesses

Every coding agent that accepts a custom base URL works. The four most common setups:

Qwen Code

export OPENAI_API_KEY="sk-…"
export OPENAI_BASE_URL="https://api.cheapweights.ai/v1"
export OPENAI_MODEL="qwen3.8-27b"

Claude Code

export ANTHROPIC_BASE_URL="https://api.cheapweights.ai"
export ANTHROPIC_AUTH_TOKEN="sk-…"
export ANTHROPIC_MODEL="qwen3.8-27b"

OpenCode

{
  "provider": {
    "cheapweights": {
      "npm": "@ai-sdk/openai-compatible",
      "options": { "baseURL": "https://api.cheapweights.ai/v1" },
      "models": { "qwen3.8-27b": {} }
    }
  }
}

Aider

export OPENAI_API_BASE=https://api.cheapweights.ai/v1
export OPENAI_API_KEY=sk-…
aider --model openai/qwen3.8-27b

Also confirmed working: Cline / Roo / Kilo ("OpenAI Compatible" provider), Continue.dev, Zed, Crush, Goose. Partial: Codex CLI (Responses API only — on our roadmap), Cursor (chat panel only). Windsurf doesn't accept custom base URLs; that's their limitation, not yours.

05 The CLI

npm i -g cheapweights
cw login                         # paste your key once
cw models                        # the price sheet in your terminal
cw ask "explain MoE routing" -m qwen3.8-27b
cw ask "write the scene" -m dolphin-3.0
cw chat                          # REPL: /model /system /balance /stream

Zero dependencies, streaming, per-request cost line on stderr. Source and README on GitHub.

06 Limits & errors