GPT-5.6 Family Guide: Sol vs Terra vs Luna — Which OpenAI Model Should You Use?

Q: Which One Should You Use?

The gpt-5.6 alias routes to Sol. This is OpenAI's most capable publicly released model, roughly corresponding to the unsuffixed flagship of earlier GPT-5 families. Best for: - Complex multi-step reasoning and math - Professional-grade coding (the gpt-5.6-sol model page highlights it for complex professional work) - Deep research, analysis, and agentic tasks - Computer-use automation where accuracy matters more than cost Skip it if: your workload is repetitive or high-volume — you're paying flags

Q: GPT-5.6 vs Previous Generations

| Model | Input $ | Output $ | Notes | |---|---|---|---| | GPT-5.6 Sol | $5.00 | $30.00 | New frontier | | GPT-5.5 | $5.00 | $30.00 | Same price, superseded | | GPT-5.6 Terra | $2.00 | $12.00 | New balanced tier | | GPT-5.4 | $2.50 | $15.00 | Superseded by Terra | | GPT-5.6 Luna | $0.20 | $1.20 | New budget tier | | GPT-5.4 mini | $0.75 | $4.50 | Superseded by Luna | The 5.6 family delivers more capability at equal or lower prices across every tier, with a 4x context jump (270K → 1.05M).

Q: Context and Long-Prompt Pricing

All GPT-5.6 models share a 1,050,000 token context window (max 922K input tokens). Watch the long-context pricing: - Prompts >272K input tokens: 2x input, 1.5x output for the whole request - Cache writes billed at 1.25x uncached input rate - Cached input is 10x cheaper ($0.50/$0.20/$0.02) For long-document workloads, prompt caching is essential — cache hits are a tenth of the input price.

Q: GPT-5.6 Sol vs Claude Opus 5 vs Grok 4.5

| Model | Input $ | Output $ | Context | Position | |---|---|---|---|---| | GPT-5.6 Sol | $5.00 | $30.00 | 1.05M | OpenAI frontier | | Claude Opus 5 | $5.00 | $25.00 | 1M | Anthropic flagship | | Claude Fable 5 | $10.00 | $50.00 | 1M | Anthropic max capability | | Grok 4.5 | $2.00 | $6.00 | 500K | xAI flagship, value pick | GPT-5.6 Sol matches Claude Opus 5's input price with a 5% larger context window; Anthropic's Fable 5 remains the raw-capability leader but at 2x the price. Grok 4.5 undercuts

⚡ Quick Answer: OpenAI's GPT-5.6 family replaces the GPT-5.5 era with three tiers: Sol ($5/$30) for complex reasoning and professional work, Terra ($2/$12) for the best intelligence-to-cost balance, and Luna ($0.20/$1.20) for cost-sensitive high-volume workloads. All three share a 1.05M token context window and 128K max output. For most developers, Terra is the sweet spot; pick Sol for hard reasoning/coding, Luna for bulk processing.

The GPT-5.6 Family at a Glance

GPT-5.6 marks a naming change for OpenAI: instead of one "GPT-5.x" flagship with mini/nano variants, the 5.6 family uses celestial tiers — Sol, Terra, Luna — each with its own identity and pricing.

Model	Role	Input $/1M	Output $/1M	Context	Max Output
GPT-5.6 Sol	Frontier, complex professional work	$5.00	$30.00	1,050,000	128K
GPT-5.6 Terra	Balanced intelligence & cost	$2.00	$12.00	1,050,000	128K
GPT-5.6 Luna	Cost-sensitive, high volume	$0.20	$1.20	1,050,000	128K

All three models:

Accept text and image input, produce text output
Support reasoning tokens, streaming, structured outputs, function calling
Have a February 16, 2026 knowledge cutoff
Are available via Responses API and Chat Completions
Support tools: web search, file search, code interpreter, computer use, MCP, and more (via Responses API)
Bill prompts >272K input tokens at 2x input / 1.5x output for the full request

Which One Should You Use?

GPT-5.6 Sol — The Frontier Workhorse

The gpt-5.6 alias routes to Sol. This is OpenAI's most capable publicly released model, roughly corresponding to the unsuffixed flagship of earlier GPT-5 families.

Best for:

Complex multi-step reasoning and math
Professional-grade coding (the gpt-5.6-sol model page highlights it for complex professional work)
Deep research, analysis, and agentic tasks
Computer-use automation where accuracy matters more than cost

Skip it if: your workload is repetitive or high-volume — you're paying flagship rates for capacity you don't need.

GPT-5.6 Terra — The Developer's Sweet Spot

Terra "balances intelligence and cost," roughly corresponding to the mini tier of earlier GPT-5 families — but with the full 1.05M context.

Best for:

Everyday coding assistants and agents
Customer-facing chat applications
Document processing at scale
Most API production workloads

At $2/$12 it undercuts GPT-5.4 ($2.5/$15) while being far more capable — the upgrade path is clear.

GPT-5.6 Luna — The Volume King

Luna "is optimized for cost-sensitive, high-volume workloads," roughly the nano tier. At $0.20/$1.20 it's OpenAI's cheapest text-capable model.

Best for:

Classification, extraction, and summarization pipelines
RAG retrieval answering
Translation and content rewriting at scale
Background batch jobs (batch API is 50% cheaper still)

GPT-5.6 vs Previous Generations

Model	Input $	Output $	Notes
GPT-5.6 Sol	$5.00	$30.00	New frontier
GPT-5.5	$5.00	$30.00	Same price, superseded
GPT-5.6 Terra	$2.00	$12.00	New balanced tier
GPT-5.4	$2.50	$15.00	Superseded by Terra
GPT-5.6 Luna	$0.20	$1.20	New budget tier
GPT-5.4 mini	$0.75	$4.50	Superseded by Luna

The 5.6 family delivers more capability at equal or lower prices across every tier, with a 4x context jump (270K → 1.05M).

Context and Long-Prompt Pricing

All GPT-5.6 models share a 1,050,000 token context window (max 922K input tokens). Watch the long-context pricing:

Prompts >272K input tokens: 2x input, 1.5x output for the whole request
Cache writes billed at 1.25x uncached input rate
Cached input is 10x cheaper ($0.50/$0.20/$0.02)

For long-document workloads, prompt caching is essential — cache hits are a tenth of the input price.

GPT-5.6 Sol vs Claude Opus 5 vs Grok 4.5

Model	Input $	Output $	Context	Position
GPT-5.6 Sol	$5.00	$30.00	1.05M	OpenAI frontier
Claude Opus 5	$5.00	$25.00	1M	Anthropic flagship
Claude Fable 5	$10.00	$50.00	1M	Anthropic max capability
Grok 4.5	$2.00	$6.00	500K	xAI flagship, value pick

GPT-5.6 Sol matches Claude Opus 5's input price with a 5% larger context window; Anthropic's Fable 5 remains the raw-capability leader but at 2x the price. Grok 4.5 undercuts both for value-oriented workloads.

Availability Note

GPT-5.6 was initially restricted by the US government in late June 2026 (the first such restriction on an OpenAI model), limited to ~20 government-vetted partners. Broad public access was promised "in the coming weeks" — the model is now rolling out via the API under standard pricing.

How to Get Started

API: use gpt-5.6 (routes to Sol), gpt-5.6-terra, or gpt-5.6-luna
Chat: GPT-5.6 powers ChatGPT web/mobile tiers
Batch: batch API at 50% of standard pricing for non-urgent workloads
Snapshots: lock in gpt-5.6-sol for stable behavior

Updated August 7, 2026 with data from OpenAI official docs. Compare live prices on our AI Model Pricing page.