OpenAI Launches GPT 5.6: Sol, Terra, and Luna

What launched

On July 9, 2026, OpenAI released the GPT 5.6 family with three tiers:

• -GPT 5.6 Sol (flagship): $5.00/$30.00 per MTok, 1M context window
• -GPT 5.6 Terra (mid): $2.50/$15.00 per MTok, 1M context window
• -GPT 5.6 Luna (fast): $1.00/$6.00 per MTok, 1M context window

All three models ship with 1,000,000 token context windows. That matches Gemini and exceeds GPT 5.5 at 128K.

How they compare across providers

Flagship tier:

• -GPT 5.6 Sol: $5.00 input, $30.00 output (reasoning 96/100)
• -Claude Opus 4.8: $5.00 input, $25.00 output (reasoning 97/100)
• -Claude Fable 5: $10.00 input, $50.00 output (reasoning 98/100)

Mid tier:

• -GPT 5.6 Terra: $2.50 input, $15.00 output (summarization 92/100)
• -GPT 5.4: $2.50 input, $15.00 output (summarization 91/100)
• -Gemini 3.1 Pro: $2.00 input, $12.00 output (summarization 87/100)
• -Claude Sonnet 5: $3.00 input, $15.00 output (summarization 91/100)

Fast tier:

• -GPT 5.6 Luna: $1.00 input, $6.00 output (classification 91/100)
• -Claude Haiku 4.5: $1.00 input, $5.00 output (classification 90/100)
• -Gemini 3 Flash: $0.50 input, $3.00 output (classification 88/100)

The standout: GPT 5.6 Terra

Terra scores 92/100 on summarization and 92/100 on QA, ranking first in both tasks. At $2.50/MTok input, it undercuts Claude Sonnet 5 by $0.50 on input pricing.

For a RAG pipeline at 6,000 input and 1,500 output tokens per call:

• -Terra: (6,000/1M x $2.50) + (1,500/1M x $15) = $0.0375/call
• -Sonnet 5: (6,000/1M x $3.00) + (1,500/1M x $15) = $0.0405/call
• -Gemini 3.1 Pro: (6,000/1M x $2.00) + (1,500/1M x $12) = $0.0300/call

Gemini 3.1 Pro remains the cheapest mid-tier option for most workloads.

Luna challenges Haiku

GPT 5.6 Luna matches Claude Haiku 4.5 at $1.00/MTok input. Luna scores 91/100 on classification versus Haiku at 90/100. Luna costs $1.00 more on output per MTok.

The tradeoff: Luna is better at classification, Haiku is cheaper on output-heavy workloads.

Compare the full 5.6 family against your current models at /teardown, or start tracking with a free dashboard.