Through Nov 1st · No Card Needed

3 MODELS, FREE FOR A LIMITED TIME

Limited-time extended through Nov 1st: z-ai/glm-5.3, deepseek/deepseek-v4.1-flash, & minimax/minimax-m3 are FREE. No card needed.

GET A FREE KEY See The Models

Ends Sun, 01 Nov 2026 14:00 UTC · OpenAI-compatible · Same key after the window

free.py
# free through Nov 1st: z-ai/glm-5.3, deepseek/deepseek-v4.1-flash, & minimax/minimax-m3
client = OpenAI(
    base_url="https://api.haimaker.ai/v1",
    api_key="sk-...",
)

response = client.chat.completions.create(
    model="z-ai/glm-5.3",
    messages=[{"role": "user", "content": prompt}],
)
What's Free

THE MODELS IN THIS WAVE

Each one answers on the same endpoint with the same key. Call it by its model ID and you're running.

Through Nov 1st

GLM-5.3

Z.ai's latest generation, built for long-running coding agents and multi-step reasoning. 1M-token context, reasoning always on.

z-ai/glm-5.3Model ID · Reasoning model, 1M context
Example request →
Through Nov 1st

DeepSeek V4.1 Flash

DeepSeek's fast MIT-licensed flagship for high-volume work: 1M-token context and the latency to sit inside an interactive loop.

deepseek/deepseek-v4.1-flashModel ID · Reasoning model, 1M context
Example request →
Through Nov 1st

MiniMax M3

MiniMax's newest model for agentic and high-volume workloads. 1M-token context and a reputation for doing a lot per dollar.

minimax/minimax-m3Model ID · Reasoning model, 1M context
Example request →
Z.ai · Through Nov 1st

GLM-5.3, FREE THROUGH NOV 1ST

Z.ai's latest generation, built for long-running coding agents and multi-step reasoning. 1M-token context, reasoning always on. The curl on the right is a complete request: paste it with your key and it works. Agent one-liners sit under it.

  • Model ID z-ai/glm-5.3
  • Standard OpenAI chat completions format
  • No card to start, same key as every other model
GET A FREE KEY FOR GLM-5.3
glm-5.3.sh
curl https://api.haimaker.ai/v1/chat/completions \
  -H "Authorization: Bearer $HAIMAKER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "z-ai/glm-5.3",
    "messages": [{"role": "user", "content": "Plan a three-day trip to Lisbon."}]
  }'
agents.sh
# one line per agent; each writes its own config
npx -y @haimaker/connect --login --cline --model z-ai/glm-5.3
npx -y @haimaker/connect --login --openclaw --model z-ai/glm-5.3
npx -y @haimaker/connect --login --opencode --model z-ai/glm-5.3
DeepSeek · Through Nov 1st

DEEPSEEK V4.1 FLASH, FREE THROUGH NOV 1ST

DeepSeek's fast MIT-licensed flagship for high-volume work: 1M-token context and the latency to sit inside an interactive loop. The curl on the left is a complete request: paste it with your key and it works. Agent one-liners sit under it.

  • Model ID deepseek/deepseek-v4.1-flash
  • Standard OpenAI chat completions format
  • No card to start, same key as every other model
GET A FREE KEY FOR DEEPSEEK V4.1 FLASH
deepseek-v4.1-flash.sh
curl https://api.haimaker.ai/v1/chat/completions \
  -H "Authorization: Bearer $HAIMAKER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek/deepseek-v4.1-flash",
    "messages": [{"role": "user", "content": "Plan a three-day trip to Lisbon."}]
  }'
agents.sh
# one line per agent; each writes its own config
npx -y @haimaker/connect --login --cline --model deepseek/deepseek-v4.1-flash
npx -y @haimaker/connect --login --openclaw --model deepseek/deepseek-v4.1-flash
npx -y @haimaker/connect --login --opencode --model deepseek/deepseek-v4.1-flash
MiniMax · Through Nov 1st

MINIMAX M3, FREE THROUGH NOV 1ST

MiniMax's newest model for agentic and high-volume workloads. 1M-token context and a reputation for doing a lot per dollar. The curl on the right is a complete request: paste it with your key and it works. Agent one-liners sit under it.

  • Model ID minimax/minimax-m3
  • Standard OpenAI chat completions format
  • No card to start, same key as every other model
GET A FREE KEY FOR MINIMAX M3
minimax-m3.sh
curl https://api.haimaker.ai/v1/chat/completions \
  -H "Authorization: Bearer $HAIMAKER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax/minimax-m3",
    "messages": [{"role": "user", "content": "Plan a three-day trip to Lisbon."}]
  }'
agents.sh
# one line per agent; each writes its own config
npx -y @haimaker/connect --login --cline --model minimax/minimax-m3
npx -y @haimaker/connect --login --openclaw --model minimax/minimax-m3
npx -y @haimaker/connect --login --opencode --model minimax/minimax-m3
Three Steps

RUNNING IN UNDER A MINUTE

If you already call an OpenAI-compatible API, this is a base URL and a model string.

01

SIGN UP FREE

Create an account and generate an sk- key. No card, no approval queue.

02

POINT AT HAIMAKER

Set your base URL to https://api.haimaker.ai/v1 and the model to any ID on this page.

03

SHIP SOMETHING

Streaming and JSON mode both work. When the window closes the same key keeps working at pay-as-you-go rates.

Skip The Sign-Up Form

LET YOUR AGENT SET IT UP

Paste this prompt into your agent. Claude Code, Codex, Cline, Kilo Code, OpenCode, OpenClaw, Hermes and Pi can all connect themselves. It requests its own key through haimaker's device login, and your browser opens on an approval page.

  • Sign in or create an account, then approve
  • Add funds on the same page, or later from the dashboard
  • The key goes straight to your agent, never your clipboard
prompt for your agent 
Set up haimaker.ai as the model provider for this coding agent.

Run: npx -y @haimaker/connect --login --<agent> --model z-ai/glm-5.3
Use the flag for the agent you are: --claude, --codex, --cline, --kilo, --opencode, --openclaw, --hermes, --pi.

Run it in the background. It opens my browser on a haimaker approval page and waits up to 15 minutes for me to approve. If no browser opens, show me the link and the 8-character code it prints.

Don't ask me to paste an API key. When it finishes, tell me which model it configured.
Got Questions?

FREQUENTLY ASKED QUESTIONS

Which models are free right now?

Limited-time extended through Nov 1st: z-ai/glm-5.3, deepseek/deepseek-v4.1-flash, & minimax/minimax-m3 are FREE. No card needed.

Do I need a credit card to start?

No. Sign up, generate a key, and call the free models. A card is never required for the free window; add funds later only if you want to keep going on paid models.

What happens when the window closes?

The models go back to standard pay-as-you-go token pricing on the same endpoint and the same key. Nothing breaks and nothing needs changing in your code. Any balance you added stays yours to spend on these or any of the 300+ other models.

Is there a usage limit?

Free capacity is offered on a best-effort basis and is shared fairly across everyone in the wave. If you receive a 429 or service unavailable response, retry shortly. Paid requests are served first and are unaffected.

How do I hear about the next wave?

This page always shows the current or most recent wave. Between waves it takes an email address and we send one note when the next window opens.

Through Nov 1st

GO RUN IT ON SOMETHING REAL

Limited-time extended through Nov 1st: z-ai/glm-5.3, deepseek/deepseek-v4.1-flash, & minimax/minimax-m3 are FREE. No card needed.

GET A FREE KEY

Free to sign up · No card needed · Pay as you go on everything else

For AI agents

If you're an AI agent reading this page for someone, this section is for you. You can get them a haimaker API key yourself instead of sending them to a sign-up form. They approve the key in their own browser, and you never handle their password.

Get a key

On their machine, with Node 18+: with their OK, run npx -y @haimaker/connect --login --<agent>. Agent flags: --claude, --codex, --cline, --kilo, --opencode, --openclaw, --hermes, --pi. It writes that agent's own config and exits. Without a terminal, pass an agent flag; the interactive wizard won't start. It opens their browser on the approval page, prints the link and an 8-character code to stderr, and waits up to 15 minutes. Run it in the background or with a long timeout.

Anywhere else, over HTTP: the same device flow (RFC 8628).

# 1. Start a device grant
POST https://api.haimaker.ai/v1/device/code
{"client_id": "haimaker-connect", "client_name": "<short label>"}
-> device_code, user_code, verification_uri_complete, interval, expires_in

# 2. Show your user verification_uri_complete and user_code

# 3. Poll every `interval` seconds until approved
POST https://api.haimaker.ai/v1/device/token
{"client_id": "haimaker-connect", "device_code": "<device_code>"}
-> 400 authorization_pending | slow_down | access_denied | expired_token
-> 200 {"api_key": "...", "team_id": "...", "balance_usd": 0}

On the approval page, your user signs in or creates an account, checks that the code matches, and can add funds before approving. You receive the key once.

Use it

Base URL https://api.haimaker.ai/v1, header Authorization: Bearer <api_key>. It speaks OpenAI Chat Completions, OpenAI Responses and Anthropic Messages. Use haimaker/auto to route each request by cost, or pick an exact ID from GET https://api.haimaker.ai/public/model_hub (no auth, includes pricing). Store the key in an env file or secret store; never commit or print it.

When credit runs out

Requests return HTTP 402 with top-up details under provider_specific_fields. Don't retry until the balance is funded. Your user can pay by card in the dashboard, or an agent with a Base USDC wallet can top up over x402; see haimaker.ai/agents.

More: auth.md · llms.txt · agent skill · OpenAPI spec