GPT-5.6 Sol, Terra & Luna: Complete Guide to OpenAI's New Model Family (2026)
What is GPT-5.6?
GPT-5.6 is OpenAI's latest model family, released on July 9, 2026. It introduces a new naming system where the number (5.6) identifies the generation, while the tier names — Sol, Terra, and Luna — identify durable capability levels that can advance on their own cadence.
| Model | Position | API Pricing (Input/Output per 1M) | Best For |
|---|---|---|---|
| Sol | Flagship | $5.00 / $30.00 | Complex coding, research, agent tasks |
| Terra | Balanced | $2.00 / $12.00 | Everyday professional work |
| Luna | Fast & Affordable | $0.20 / $1.20 | High-volume, cost-sensitive tasks |
Price update (July 30, 2026): OpenAI cut Luna prices by 80% and Terra by 20%, citing efficiency improvements and competitive pressure. The prices above reflect the new rates.
Specifications
| Specification | Sol | Terra | Luna |
|---|---|---|---|
| Context Window | 1M tokens | 1M tokens | 1M tokens |
| Max Output | 128K tokens | 128K tokens | 128K tokens |
| Knowledge Cutoff | Feb 16, 2026 | Feb 16, 2026 | Feb 16, 2026 |
| Reasoning Efforts | none, low, medium, high, xhigh, max | none, low, medium, high | none, low, medium |
| Ultra Mode (sub-agents) | ✅ | ❌ | ❌ |
| API Pricing (input) | $5.00/M | $2.00/M | $0.20/M |
| API Pricing (output) | $30.00/M | $12.00/M | $1.20/M |
| Cerebras Speed | Up to 750 tok/s | — | — |
Prompt Caching
All three models support prompt caching with a 30-minute minimum cache life and explicit cache breakpoints:
| Cache Type | Pricing |
|---|---|
| Cache Write | 1.25x uncached input rate |
| Cache Read | 90% discount on uncached input rate |
Luna: The 80% Price Cut Story
The biggest news in the GPT-5.6 family is Luna's dramatic price reduction on July 30, 2026:
| Metric | Before July 30 | After July 30 | Change |
|---|---|---|---|
| Luna Input | $1.00/M | $0.20/M | -80% |
| Luna Output | $6.00/M | $1.20/M | -80% |
| Terra Input | $2.50/M | $2.00/M | -20% |
| Terra Output | $15.00/M | $12.00/M | -20% |
OpenAI attributed the cuts to a 20% reduction in serving costs and 15% improvement in token-generation efficiency. Media reports also point to competitive pressure from Chinese AI labs like DeepSeek and ByteDance, which have been aggressively pricing their models.
At $0.20/$1.20 per 1M tokens, Luna is now OpenAI's most affordable model ever — competitive with DeepSeek V4 Flash ($0.14/$0.28) while offering access to OpenAI's ecosystem.
Sol Fast Mode
For users who need speed, Sol now offers a Fast mode at 2x the standard price, delivering up to 2.5x faster generation. This is available through the API and Codex, and is particularly useful for interactive coding sessions.
Capabilities
Sol: The Flagship
GPT-5.6 Sol is OpenAI's most capable model, designed for complex tasks across coding, research, cybersecurity, and agentic workflows.
Key benchmark results:
| Benchmark | Sol Score | Notes |
|---|---|---|
| Agents' Last Exam | 53.6 | Beats Claude Fable 5 by 13.1 points |
| Terminal-Bench 2.1 | State of the Art | Command-line workflow tasks |
| SWE-Bench Pro | 64.6% | Trails Fable 5 (80%), but OpenAI disputes benchmark reliability |
| GeneBench v1 | Strong | Better than GPT-5.5 with fewer tokens |
| ExploitBench | Competitive with Mythos Preview | Using ~1/3 of the output tokens |
Sol Pro is a separate model variant available in ChatGPT for more difficult, longer-running workflows. It extends Sol's capabilities for tasks that benefit from additional compute.
Max reasoning effort (levels: none through max) gives Sol more time to reason deeply on complex problems. The "max" level is the most thorough but also the most expensive.
Ultra mode goes beyond a single agent by leveraging sub-agents to accelerate complex work. This is available in the ChatGPT interface and represents OpenAI's most powerful configuration.
Terra: The Balanced Choice
Terra is designed for everyday professional work. At $2/$12 per 1M tokens (after the 20% price cut), it offers competitive performance with GPT-5.5 at half the price. It supports reasoning efforts up to "high" and is suitable for:
- Content creation and editing
- Data analysis and visualization
- Code generation and review
- Research synthesis
Luna: The Budget Champion
Luna is the fast and affordable option. At $0.20/$1.20 per 1M tokens, it's 80% cheaper than its launch price and competitive with the most affordable models on the market. It supports reasoning efforts up to "medium" and excels at:
- High-volume text processing
- Simple coding tasks
- Classification and extraction
- Customer support automation
- Batch processing
New API Features
GPT-5.6 introduces several new capabilities through the API:
Programmatic Tool Calling
Models can compose and run JavaScript that orchestrates tool calls. This bridges the gap between individual tool calls and full terminal sessions, enabling more complex autonomous workflows.
Multi-Agent
The API now supports spinning up sub-agents for parallel, focused work. This is a significant step toward practical multi-agent systems, baked directly into the core API rather than requiring custom orchestration.
Prompt Cache Breakpoints
Inspired by Claude's approach, developers can now explicitly set cache breakpoints rather than relying on automatic detection. Combined with a 30-minute minimum cache life, this gives developers more control over caching costs.
Image Detail Control
The new detail: original option on image requests avoids resizing images before processing, preserving maximum visual information for analysis tasks.
GPT-5.6 in ChatGPT
In the ChatGPT interface, the three models map to different user experiences:
| ChatGPT Plan | Models Available | Key Features |
|---|---|---|
| Free | GPT-4o (not GPT-5.6) | Basic access |
| Plus | GPT-5.6 Sol, GPT-5.6 Terra | Up to medium reasoning |
| Pro | GPT-5.6 Sol Pro, GPT-5.6 Sol, GPT-5.6 Terra | Max reasoning, Ultra mode |
Sol Pro is a ChatGPT-specific variant that extends Sol's capabilities for demanding workflows. Max is a reasoning effort slider (none → max). Ultra activates sub-agent mode for complex multi-step tasks.
Competitor Comparison
| Aspect | GPT-5.6 Sol | Claude Fable 5 | DeepSeek V4 Flash | Qwen 3.8-Max |
|---|---|---|---|---|
| Input Price / 1M | $5.00 | ~$10.00 | $0.14 | $2.00 |
| Output Price / 1M | $30.00 | ~$50.00 | $0.28 | $6.00 |
| Context | 1M | 200K | 1M | 1M |
| Agents' Last Exam | 53.6 | ~40.4 | — | — |
| SWE-Bench Pro | 64.6% | 80% | — | — |
| Open Weights | ❌ | ❌ | ✅ MIT | ✅ Promised |
GPT-5.6 Sol is the most expensive model here, but it leads on agentic benchmarks. Luna, at $0.20/$1.20, is the most affordable entry point into the GPT-5.6 ecosystem and directly competes with DeepSeek V4 Flash on price.
Who Should Use Which Model?
| User Profile | Recommended Model | Rationale |
|---|---|---|
| Agent developers | Sol | Best agentic performance, Terminal-Bench SOTA |
| Professional developers | Sol or Terra | Sol for complex tasks, Terra for daily work |
| High-volume API users | Luna | $0.20/$1.20 — best value in OpenAI's lineup |
| Content creators | Terra | Balanced performance and cost |
| Budget-conscious teams | Luna or DeepSeek V4 Flash | Luna for OpenAI ecosystem, V4 Flash for open source |
| Enterprise | Sol Pro | Extended capabilities, dedicated support |
Summary
GPT-5.6 represents OpenAI's most comprehensive model family to date. Sol sets new benchmarks in agentic performance, Terra offers a balanced middle ground, and Luna — now 80% cheaper — is OpenAI's most affordable model ever.
The pricing landscape has shifted dramatically. Luna at $0.20/$1.20 directly competes with DeepSeek V4 Flash and other low-cost models, while Sol maintains OpenAI's position at the frontier of agentic capabilities. The new API features — Programmatic Tool Calling, Multi-agent support, and Prompt Cache Breakpoints — give developers tools that weren't available before.
For users already in the OpenAI ecosystem, the GPT-5.6 family offers a clear upgrade path. For new users, the choice between Sol, Terra, and Luna depends on the balance of capability and cost your workflow demands.
Frequently Asked Questions
What are the three GPT-5.6 models?+
How much does GPT-5.6 Luna cost after the price cut?+
What is the difference between Sol Pro, Max, and Ultra?+
How does GPT-5.6 Sol compare to Claude Fable 5?+
Is GPT-5.6 available to everyone?+
Pros
- Three capability tiers covering every use case and budget
- Luna is now the best value in OpenAI's lineup at $0.20/$1.20 per 1M tokens (80% price cut)
- Sol sets new SOTA on Agents' Last Exam (53.6) and Terminal-Bench 2.1
- 1M token context window across all three models
- New API features: Programmatic Tool Calling, Multi-agent, Prompt Cache Breakpoints
- Sol available on Cerebras at up to 750 tokens/sec
Cons
- Sol is expensive at $5/$30 per 1M tokens
- Still rolling out gradually — not available to all ChatGPT users yet
- SWE-Bench Pro score (64.6%) trails Claude Fable 5 (80%)
- Knowledge cutoff is Feb 16, 2026 — not the most up-to-date
- Sol Pro and Ultra mode only available in ChatGPT, not API
Related Articles
Qwen 3.8-Max: Alibaba's 2.4 Trillion Parameter Open-Source Flagship (2026)
Qwen 3.8-Max is here — 2.4T parameters, 1M context, open weights coming next week. Full review covering autonomous coding, agent benchmarks, pricing, and how it compares to GPT-5.6 and Fable 5.
Kimi K3: Moonshot's 2.8T Frontier AI Model — Complete Review 2026
In-depth review of Kimi K3, Moonshot AI's 2.8-trillion-parameter open-weight model. Features, benchmarks, pricing ($3/$15 per MTok), and how it compares to GPT-5.5, Claude Opus 4.8, and DeepSeek.
DeepSeek V4 Flash: Public Beta Launches With Major Agent Performance Boost (2026)
DeepSeek V4 Flash API is now in public beta. 284B MoE model, 1M context, 384K output, and dramatically improved agent benchmarks. Full review with pricing, benchmarks, and use cases.
Grok 4.5: Complete Guide to xAI's Coding-Focused Frontier Model (2026)
In-depth review of Grok 4.5—features, pricing, benchmarks, and real-world use cases. Learn how xAI's coding-focused model compares to Claude, GPT, and Gemini.