GPT-5.6 Luna Is Now Free on Neosantara

GPT-5.6 Luna is now available on Neosantara — at zero cost. The model is served on Kiro and managed by Neosantara, accessible to all users including the Free tier with a 100% discount on usage fees.

GPT-5.6 Luna is the efficient variant of OpenAI's GPT-5.6 family released in July 2026. It's designed for text generation, reasoning, and function calling with solid performance across a range of tasks (OpenAI, 2026).

#What Is GPT-5.6 Luna?

GPT-5.6 Luna is a text-only model from the GPT-5.6 family, optimized for speed and cost efficiency. "Luna" is the fastest and most affordable variant in the family — below Sol (flagship) and Terra (balanced). All GPT-5.6 models share a 1 million token context window and a February 2026 knowledge cutoff.

Good use cases include:

• Text and content generation
• Coding assistance and code generation
• Multi-step reasoning with reasoning_effort control
• Autonomous tool calling in agentic workflows

#GPT-5.6 Luna Capabilities

GPT-5.6 Luna supports autonomous tool calling — the model independently decides when and which tools to invoke. However, tool_choice cannot be forced; the model retains control over whether a tool is used.

The reasoning_effort parameter lets you control how deeply the model thinks before responding. Available options: low, medium, and high.

#Free Tier Access and 100% Discount

GPT-5.6 Luna is available to all Neosantara users, including the Free tier. The applicable discount is 100%, meaning the usage cost charged to users is Rp0.

Here's the official base pricing (before discount) for transparency:

Important to understand: "free" here means the usage cost after a 100% discount is Rp0. It doesn't mean unlimited. Rate limits still apply based on your account tier, and token usage is still recorded for rate limiting and analytics purposes.

Caching is not available on this route, so cache read/write pricing does not apply.

#Use with Coding Assistants

The primary use case for GPT-5.6 Luna on Neosantara is as a backend model for coding assistants. Two ways to access it:

Option 1: PAYG Key (Free for Luna) — Use a regular API key on the /v1 endpoint. GPT-5.6 Luna has a 100% discount, so cost is Rp0.

Option 2: Coding Plan — Flat-fee subscription with prompt-based quota on the dedicated /coding/v1 endpoint. Best if you also use other models (Claude, Gemini) beyond Luna.

Here's the setup per tool using a PAYG key (free for Luna):

Claude Code — Edit ~/.claude/settings.json:

{
  "env": {
    "OPENAI_API_KEY": "nsk_YOUR_KEY",
    "OPENAI_BASE_URL": "https://api.neosantara.xyz/v1",
    "OPENAI_MODEL": "gpt-5.6-luna"
  }
}

Codex CLI — Set environment variables:

export OPENAI_API_KEY="nsk_YOUR_KEY"
export OPENAI_BASE_URL="https://api.neosantara.xyz/v1"
codex --model gpt-5.6-luna

Cline (VS Code) — In extension settings:

API Provider: OpenAI Compatible
Base URL:     https://api.neosantara.xyz/v1
API Key:      nsk_YOUR_KEY
Model:        gpt-5.6-luna

Cursor — Settings → Models → Add Custom Model:

OpenAI Base URL: https://api.neosantara.xyz/v1
API Key:         nsk_YOUR_KEY
Model name:      gpt-5.6-luna

Continue / OpenCode — Config file or UI:

Base URL: https://api.neosantara.xyz/v1
API Key:  nsk_YOUR_KEY
Model:    gpt-5.6-luna

All of the above work because Neosantara exposes a /v1/chat/completions endpoint fully compatible with the OpenAI format — including streaming, function calling, and reasoning. Full guide at docs.neosantara.xyz/en/coding-plan/tool-integration.

#Limitations You Should Know

Transparency matters. Here are the GPT-5.6 Luna limitations on Neosantara you should know before integrating:

The model is text-only. No vision or image input support. Don't send images or image URLs in message content.

Some parameters are ignored. The following parameters can be sent to the API but won't affect output:

• tool_choice — the model decides autonomously
• response_format: {"type": "json_object"} — native JSON mode is not available
• top_p
• frequency_penalty
• presence_penalty
• logit_bias
• seed

Only one output choice. Setting n > 1 still produces a single choice. If you need multiple completions, make separate requests.

No deterministic sampling. Since seed is ignored, outputs cannot be reproduced identically across requests.

Subject to Neosantara PAYG rate limits. Even though usage cost is Rp0, the Free tier is still bound by standard throughput limits (RPM, ITPM, OTPM). If exceeded, the API returns 429 Too Many Requests. See the rate limit documentation for per-tier details.

#API Usage Example

GPT-5.6 Luna is accessed via Neosantara's standard OpenAI-compatible endpoint. Here's a cURL example:

curl https://api.neosantara.xyz/v1/chat/completions \
  -H "Authorization: Bearer $NEOSANTARA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.6-luna",
    "messages": [
      {
        "role": "user",
        "content": "Write a JavaScript function to validate an email address."
      }
    ],
    "reasoning_effort": "high"
  }'

You can use any SDK compatible with the OpenAI API — just point base_url to https://api.neosantara.xyz/v1 and use your Neosantara API key.

Example with Python (openai SDK):

from openai import OpenAI

client = OpenAI(
    api_key="NEOSANTARA_API_KEY",
    base_url="https://api.neosantara.xyz/v1"
)

response = client.chat.completions.create(
    model="gpt-5.6-luna",
    messages=[
        {"role": "user", "content": "Explain how garbage collection works in Go."}
    ],
    reasoning_effort="medium"
)

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

#Get Started

GPT-5.6 Luna is live and ready to use. No tier upgrade or special setup required.

1. Get your API key — Open the Neosantara dashboard and create or copy your API key
2. Send your first request — Use the examples above with model gpt-5.6-luna
3. Read the docs — Full documentation at docs.neosantara.xyz

#FAQ

Is GPT-5.6 Luna really free?

Usage cost after the 100% discount is Rp0. Token usage is still tracked and rate limits still apply based on your account tier. No hidden fees.

Can I use it in production?

Yes. The model is available via Neosantara's standard API with SLA matching your account tier. However, note the limitations listed above, especially regarding ignored parameters.

What about rate limits?

Rate limits follow your Neosantara account tier. The Free tier has specific limits on RPM (requests per minute) and token throughput. Check the rate limit documentation for per-tier details.

Does it support streaming?

Yes. Add "stream": true to the request body to receive streaming responses.

What's the difference between Luna, Sol, and Terra?

Sol ($5/$30 per 1M tokens) is the flagship for reasoning and agentic work. Terra ($2.50/$15) is the balanced everyday production model. Luna ($1/$6) is the fastest and cheapest, built for high-volume workloads. All three are distilled from the same training run and share a 1 million token context window. Currently only Luna is available on Neosantara.