What are GPT-5.6 Sol, Terra, and Luna?

Sol, Terra, and Luna are the three tiers of OpenAI's GPT-5.6 family, publicly available since July 9, 2026. Sol ($5/$30 per MTok) is the flagship for complex coding, cybersecurity, and long-horizon workflows. Terra ($2.50/$15) is the balanced mid-tier for everyday coding and business tasks. Luna ($1/$6) is the fastest and cheapest — designed for high-volume classification, summarization, and routing.

Which GPT-5.6 tier should I use for coding?

For most coding work, Terra ($2.50/$15 per MTok) delivers roughly 90% of Sol's quality at half the price. Use Sol for hardest tasks — long multi-file refactors, terminal workflows requiring Ultra mode, or cybersecurity vulnerability research. Use Luna for simple autocomplete, batch code review, or low-stakes generation.

Is GPT-5.6 available in ChatGPT?

Yes, as of July 9, 2026 — the public launch expanded GPT-5.6 from the API-and-Codex preview to ChatGPT Plus, Pro, and Enterprise. During the preview (June 26 to July 9, 2026) it was API- and Codex-only. Plus users get Terra by default with rate-limited Sol access; Pro and Enterprise get full Sol and Ultra access.

Which GPT-5.6 tier is cheapest?

GPT-5.6 Luna at $1 per million input tokens and $6 per million output tokens is the cheapest tier. It's roughly 5x cheaper than Sol and half the price of Terra. Luna is designed for high-volume, easy-to-verify work: summaries, classification, routing, autocomplete, and structured extraction.

Quick Answer

GPT-5.6 Sol vs Terra vs Luna: Which OpenAI Tier to Pick (July 2026)

Published: July 9, 2026

GPT-5.6 Sol vs Terra vs Luna at a Glance

OpenAI publicly released the GPT-5.6 family — Sol, Terra, and Luna — on Thursday, July 9, 2026, ending the API-only preview that started June 26. Naming pattern: solar-system objects, with capability decreasing and speed increasing as you move from Sun (Sol) → Earth (Terra) → Moon (Luna).

Tier	Input / MTok	Output / MTok	Speed	Positioning
GPT-5.6 Sol	$5.00	$30.00	Standard	Flagship — hardest tasks
GPT-5.6 Terra	$2.50	$15.00	Fast	Mid-tier — balanced
GPT-5.6 Luna	$1.00	$6.00	Very fast	Cheap volume workhorse
Sol Ultra	$12.50	$75.00	Up to 750 tok/s	Multi-agent Sol variant

When to Use Each Tier

Sol — Flagship

Use for:

Complex software engineering (multi-file refactors, migrations)
Cybersecurity vulnerability research (Sol’s specialty)
Long-horizon coding (Ultra mode: 91.9% on Terminal-Bench 2.1)
Scientific workflows: genomics, quantitative biology
Any task where model quality dominates cost

Don’t use for: Chat, autocomplete, batch classification — Sol is overkill and 5x pricier than Luna.

Terra — Mid-Tier Workhorse

Use for:

Everyday coding assistance (Cursor, Copilot backends)
Business analysis, drafting, research summaries
Agent workflows that don’t need Sol’s specialization
Anything where Sonnet 5 would work but you want an OpenAI equivalent

Positioning: Terra roughly matches GPT-5.5 quality at ~50% of GPT-5.5’s original price. It’s the direct competitor to Claude Sonnet 5 ($2/$10 intro / $3/$15 standard).

Luna — Cheap Volume

Use for:

High-volume chat (support bots, FAQ answering)
Classification, tagging, sentiment
Batch summarization
Routing (deciding which more expensive model to call)
Autocomplete and inline code suggestions

Positioning: Luna at $1/$6 competes with Claude Haiku 4.5 ($0.80/$4 as of July 2026) and DeepSeek V4 Flash ($0.15/$0.60). It’s not the absolute cheapest, but it’s the cheapest frontier-lab option with reliable quality.

Head-to-Head vs Claude

Task	Cheapest Anthropic	Cheapest GPT-5.6	Winner (July 2026)
Classification	Haiku 4.5 ($0.80/$4)	Luna ($1/$6)	Haiku on price
Everyday coding	Sonnet 5 ($2/$10 intro)	Terra ($2.50/$15)	Sonnet on price and 1M context
Hardest coding	Fable 5 ($10/$50, gated)	Sol Ultra ($12.50/$75)	Ultra for public availability
Deep analysis	Opus 4.8 ($15/$75)	Sol ($5/$30)	Sol on price, Opus on nuance

Decision Framework

What's the task?
│
├── Volume + speed (classification, routing, chat)
│   → Luna ($1/$6)  — or Haiku 4.5 for slightly cheaper
│
├── Everyday coding + business tasks
│   → Terra ($2.50/$15)  — or Sonnet 5 ($2/$10 intro) for 1M context
│
├── Complex coding, cybersecurity, long horizon
│   → Sol ($5/$30)
│
├── Multi-agent orchestration + hardest terminal/coding
│   → Sol Ultra ($12.50/$75)
│
└── Deep writing nuance, creative long-form
    → Claude Opus 4.8 (still better than any GPT-5.6 tier here)

ChatGPT Access (July 9, 2026)

Plan	GPT-5.6 Access
Free	Luna with tight rate limits
Plus ($20/mo)	Terra by default, limited Sol
Pro ($200/mo)	Full Sol, limited Ultra
Business/Enterprise	Full Sol and Ultra, admin controls
API	All tiers pay-as-you-go

The government-gating from the preview period is lifted for Sol and Terra as of July 9. Sol Ultra remains under a heavier safety-testing regime for high-risk queries (bio, cyber-offense) — expect refusals in those domains.

The Bottom Line

For July 2026, the smart routing pattern for OpenAI’s stack is:

Luna for volume — chat, classification, cheap generation
Terra for everyday work — the new default for most coding and business tasks
Sol only when Terra isn’t enough — hardest coding, cyber research, scientific workflows
Sol Ultra only when you need the multi-agent boost — long-horizon tasks that would exhaust a single context

Most teams over-buy on Sol. Start with Terra and only escalate to Sol on tasks where you can measure the quality gap.

Sources

OpenAI: Previewing GPT-5.6 Sol — official June 26, 2026 announcement
Nextgov: GPT-5.6 Public Availability — July 9 launch coverage
Coursiv: GPT-5.6 Terra Pricing and Model IDs — tier breakdown
Coursiv: GPT-5.6 Luna Overview — Luna specifications