GPT-5.6 Luna is OpenAI’s cost-efficient GPT-5.6 tier: a million-token API model for coding, agents and long-context work when Sol and Terra are more than the job needs.
On 30 July 2026 OpenAI cut Luna’s standard API rates by 80%. At $0.20 input and $1.20 output per million tokens it is now the family’s default volume tier, not merely a cheaper sibling.
At a glance
| Fact | Detail |
| --- | --- |
| Provider | OpenAI |
| Type | Commercial |
| Best for | Coding, agents, research, large document analysis |
| Context window | 1,050,000 tokens |
| API available | Yes |
| Local use | No |
| Last checked | 31 August 2026 |
Our verdict
Luna is the GPT-5.6 default for production routing. The July cut makes million-token context cheap enough for high-volume coding and research, while Sol remains the pick for the hardest jobs. It is still a hosted OpenAI model: do not use it if you need local weights.
| Score | Rating |
| --- | --- |
| Overall | ★★★★☆ |
| Coding | ★★★★☆ |
| Agents | ★★★★☆ |
| Research | ★★★★☆ |
| Budget | ★★★★★ |
Pricing
Standard API prices per million tokens, short context:
| Meter | Price |
| --- | --- |
| Input | $0.20 / million tokens |
| Cached input | $0.02 / million tokens |
| Cache writes | $0.25 / million tokens |
| Output | $1.20 / million tokens |
Long-context rates (OpenAI’s long-context columns on the official pricing table) are $0.40 input, $0.04 cached input and $1.80 output per million tokens. Batch and Fast mode differ from these standard rates.
Best for
- High-volume coding, agents and research
- Long-context analysis where Sol is unnecessary
- Tool-using workflows with deliberate model routing
Avoid when
- You need local inference or open weights
- The task needs Sol-class reasoning
- You cannot tolerate a hosted-provider dependency
What changed
- 31 August 2026: Official pricing re-checked. Standard Luna rates are $0.20 / $0.02 / $1.20 per million tokens.
- 30 July 2026: OpenAI reduced GPT-5.6 Luna pricing by 80%.
- 17 July 2026: Profile created for GPT-5.6 Luna.
Related news
- OpenAI’s GPT-5.6 push is really a pricing war for enterprise AI
- SpaceXAI ships Grok 4.5, Musk calls it an Opus-Class model at half the price
- Alibaba bans employees from using Anthropic’s Claude Code
