GPT-5.6 Sol
OpenAILarge Language ModelProprietaryFlagship tier of the GPT-5.6 family (Sol/Terra/Luna), released July 2026. State-of-the-art agentic coding with unusually low token usage, strong cybersecurity defense and science capabilities, and a 1.05M token context window.
Abilities
Use Cases
Available in Tools
Availability
How to Use
Pros
- State-of-the-art agentic coding — top of Artificial Analysis Coding Agent Index with under half the output tokens of competitors
- 1.05M token context window with 128K max output across all three tiers
- Token efficiency lowers real cost — fewer output tokens per task than prior GPT flagships
- Family covers all price points — Terra and Luna tiers share the context window at a fraction of the cost
- Promotional pricing of $4/$20 per MTok through ~Nov 2026 (list $5/$30)
Cons
- Closed source — no model weights or training transparency
- Already shadowed by GPT-6 Astra (Sept 2026) at the very top of the OpenAI lineup
- Promotional pricing is temporary — list price returns to $5/$30 per MTok
What to Use It For
Perfect For
SOTA coding-agent benchmark results combined with unusually low token usage per task
Officially highlighted strength of the 5.6 family, with a dedicated restricted Cyber variant for approved partners
1.05M context fits entire codebases and large document sets in one request
Good For
Strong science and multi-step reasoning capabilities per the official launch materials
Route easy traffic to Luna and hard tasks to Sol without changing context limits
Not Recommended
Luna at $0.20/$1.20 or GPT-4.1 Nano are far cheaper for simple high-volume tasks
Try instead: GPT-4.1 Nano, Claude Haiku 4.5
Closed source with no available weights — completely dependent on OpenAI API
Try instead: Llama 4 Maverick, Qwen 3
Do Not Use For
LLM — text and vision understanding only, no image creation
Try instead: GPT Image 2.5, FLUX.2 Pro
No video output — and OpenAI has no video model since Sora shut down
Try instead: Veo, Kling 2.5 Turbo Pro