GPT-5.6 Luna versus Muse Spark 1.3

What are we comparing?

This comparison pits two budget models, built for high volumes at a low price against each other: GPT-5.6 Luna and Muse Spark 1.3, one from OpenAI, the other from Meta. The first dates from July 9, 2026, the second from September 2, 2026.

All figures below come from public sources checked on October 22, 2026: models.dev and OpenRouter for prices and capabilities, Epoch AI for scores.

Performance

On our capability index, GPT-5.6 Luna scores 79.6 out of 100 and Muse Spark 1.3 82.7. Muse Spark 1.3 leads by 3.1 points.

On the 3 benchmarks reported for both, GPT-5.6 Luna wins 1 time and Muse Spark 1.3 2:

• OTIS Mock AIME (competition mathematics problems): GPT-5.6 Luna 98.3%, Muse Spark 1.3 99.2%
• FrontierMath (very hard research-level mathematics): GPT-5.6 Luna 82.1%, Muse Spark 1.3 74.4%
• APEX-Agents (professional tasks carried out by an agent): GPT-5.6 Luna 43%, Muse Spark 1.3 58.6%

Pricing

GPT-5.6 Luna costs $0.20 per million input tokens and $1.2 per million output tokens; Muse Spark 1.3 costs $0.10 and $0.20.

For a typical agent task (read 20,000 tokens, write 2,000), expect $0.006 with GPT-5.6 Luna versus $0.002 with Muse Spark 1.3. Muse Spark 1.3 is 2.7 times cheaper, a gap that adds up quickly over 1,000 tasks: $4 difference.

Run the numbers with your own volumes in the agent cost calculator.

Under the hood

GPT-5.6 Luna reads more in one go: 1,050,000 tokens of context (roughly 1,575 pages), versus 1,048,576 for Muse Spark 1.3.

Reasoning mode: GPT-5.6 Luna yes, Muse Spark 1.3 yes. Tool calling: GPT-5.6 Luna yes, Muse Spark 1.3 yes. Image input: GPT-5.6 Luna yes, Muse Spark 1.3 yes.

Neither publishes its weights: both are used through an API.

In short

Our reading of the data:

• Choose Muse Spark 1.3 for a higher capability index, a lower cost and a more recent release.

This verdict relies on public figures; it does not replace a trial on your own tasks.