Models

Live

Qwen-SEA-LION-v4.5-27B-IT

A 27B instruction-tuned model from AI Singapore's SEA-LION family, built for Southeast Asian languages alongside English. AI Singapore is a MarsCompute collaborator.

Model ID

qwen-sea-lion-v4.5-27b

Developer

AI Singapore

Interface

Chat completions, text in and text out

Streaming

Yes

Price

$0.60 input and $3.60 output per 1M tokens

Good to know

The model's reasoning output is switched off by default, so replies are direct and you are charged for the answer only.

curl https://api.marscompute.ai/v1/chat/completions \
  -H "Authorization: Bearer $MARSCOMPUTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen-sea-lion-v4.5-27b",
    "messages": [
      {"role": "user", "content": "Say hello in Malay."}
    ],
    "max_tokens": 256
  }'

Live

GPT-6 Astra

OpenAI's GPT-6 Astra, available with the same keys, usage records and spending limits as every other model on MarsCompute.

Model ID

gpt-6-astra

Developer

OpenAI

Interface

Chat completions, text in and text out

Streaming

Yes

Price

$10.00 input and $50.00 output per 1M tokens

Good to know

Only the default temperature is accepted. A request that sets another value is rejected with a 400 error.

curl https://api.marscompute.ai/v1/chat/completions \
  -H "Authorization: Bearer $MARSCOMPUTE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-6-astra",
    "messages": [
      {"role": "user", "content": "Write a haiku about Mars."}
    ],
    "max_tokens": 256
  }'

Coming soon

Kimi K3

An open-weight model from Moonshot AI, to be served on GPU capacity operated by MarsCompute, where our optimization stack applies.

Coming soon

GLM-5.3

An open-weight model from Z.ai, to be served on GPU capacity operated by MarsCompute, where our optimization stack applies.