OpenAI Unveils GPT-5.6 Series: Sol, Terra, and Luna Launch in Limited Preview

OpenAI New Models
GPT-5.6 series — Sol (flagship), Terra (balanced), Luna (lightweight) — now available for trusted partners.

🚀 Key Announcement

OpenAI initiated a limited preview of its new GPT-5.6 model series on June 26, 2026, delivering three distinct variants optimized for different use cases:

Model	Target Use Case	Key Strengths
Sol	Flagship & advanced tasks	Coding, bio workflows, cybersecurity, long-horizon agent reasoning
Terra	Daily & high-frequency workloads	Near-GPT-5.5 performance at ~50% lower cost
Luna	Cost-sensitive scale deployments	Fastest inference, lowest price per token

💡 Sol introduces max reasoning depth for extended deliberation and ultra mode—orchestrating multiple sub-agents for complex multi-step tasks.

⚙️ Technical Enhancements

Predictable Prompt Caching: Explicit cache breakpoints + minimum 30-minute cache lifetime.
Enhanced Security Stack: Sol features OpenAI’s strongest built-in safety layer, including real-time output monitoring, account-level risk signals, differential access controls, and continuous red-teaming.
Risk Assessment: Rated High (not Critical) for cybersecurity and bio/chem-related risks per official GPT-5.6 Preview System Card.
Red-Team Investment: >700,000 A100-equivalent GPU hours dedicated to automated adversarial testing.

💰 Pricing (per 1M tokens)

Model	Input	Output
Sol	$5.00	$30.00
Terra	$2.50	$15.00
Luna	$1.00	$6.00

✅ All tiers significantly undercut Anthropic’s Opus/Fable pricing — Luna is notably competitive with leading domestic open-weight models.

🌐 Availability & Rollout Plan

Initial Access: API and Codex only — restricted to vetted enterprise partners and research institutions.
Regulatory Alignment: Pre-release briefing conducted with U.S. government agencies; rollout reflects coordinated safety governance protocols.
Phased Expansion: Planned release to ChatGPT, Codex, and broader API access within the coming weeks.

🎯 Strategic Focus

GPT-5.6 marks a deliberate pivot from pure conversational fluency toward robust agentic capability, emphasizing:
– End-to-end code generation & vulnerability remediation (not exploitation)
– Domain-specialized reasoning (biology, security, systems engineering)
– Composable intelligent workflows with deterministic caching & orchestration
– Responsible scaling via controlled preview, rigorous red-teaming, and policy-aligned deployment

Source: OpenAI Official Blog — “Previewing GPT-5.6 Sol: A Next-Generation Model” & GPT-5.6 Preview System Card