High-volume, cost-sensitive work
Choose GPT-5 nano
GPT-5 nano costs 31% less on a typical mix of three input tokens for every output token ($0.14 against $0.20 per million tokens).
GPT-5 nano vs GPT-6 Luna: price, context and capabilities
Choose GPT-5 nano for high-volume, cost-sensitive work, writing and code generation and reused prompts and documents (RAG, agents). Choose GPT-6 Luna for very long inputs, reasoning over long documents, the latest release and the most recent knowledge.
In short
- GPT-5 nano is 31% cheaper than GPT-6 Luna at the headline price ($0.1375 / 1M tokens vs $0.20 / 1M tokens).
- GPT-6 Luna reads more per request: 1.05M tokens against 400k.
- GPT-6 Luna is the newer model (Sep 22, 2026, against Aug 7, 2025).
Which should you choose?
Each answer below comes straight from the published prices, specifications and independent benchmark results.
Writing and code generation
Choose GPT-5 nano
GPT-5 nano charges $0.40 per million output tokens, against $0.50 for GPT-6 Luna. Output dominates the bill when answers are long.
Reused prompts and documents (RAG, agents)
Choose GPT-5 nano
GPT-5 nano charges <$0.01 per million cached input tokens, against $0.01 for GPT-6 Luna.
Very long inputs
Choose GPT-6 Luna
GPT-6 Luna reads up to 1.05M tokens per request, against 400k for GPT-5 nano.
Reasoning over long documents
Choose GPT-6 Luna
GPT-6 Luna scores 83.3% on Long-context reasoning (AA-LCR), against 45% for GPT-5 nano.
Latest release
Choose GPT-6 Luna
GPT-6 Luna is the newer release (Sep 22, 2026, against Aug 7, 2025 for GPT-5 nano).
Most recent knowledge
Choose GPT-6 Luna
GPT-6 Luna was trained on data up to May 2026, against May 2024 for GPT-5 nano.
Monthly cost at three volumes
| Volume | GPT-5 nano | GPT-6 Luna | Difference |
|---|---|---|---|
| Prototype: 1M input, 250K output tokens | $0.15 | $0.23 | GPT-5 nano, $0.08 less |
| Production: 20M input, 5M output tokens | $3.00 | $4.50 | GPT-5 nano, $1.50 less |
| Scale: 500M input, 100M output tokens | $65.00 | $100.00 | GPT-5 nano, $35.00 less |
Standard rates, without caching or batch discounts. Volumes are per month.
Specifications side by side
| GPT-5 nano | GPT-6 Luna | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Category | Language models | Language models |
| Status | Legacy | Generally available |
| Headline price | $0.1375 / 1M tokens | $0.20 / 1M tokens |
| Input | $0.05 / 1M tokens | $0.10 / 1M tokens |
| Output | $0.40 / 1M tokens | $0.50 / 1M tokens |
| Cached input | $0.005 / 1M tokens | $0.01 / 1M tokens |
| Context window | 400k | 1.05M |
| Max output | 128k | 128k |
| Released | Aug 7, 2025 | Sep 22, 2026 |
| Knowledge cutoff | May 2024 | May 2026 |
| Licence | Proprietary | Proprietary |
| Capabilities | Vision, Reasoning, Tool calling, Structured output, Prompt caching, Batch API | Vision, Reasoning, Tool calling, Structured output, Prompt caching, Batch API |
| API model id | gpt-5-nano | gpt-6-luna |
Benchmarks side by side
Leader in bold.
| Benchmark | GPT-5 nano | GPT-6 Luna |
|---|---|---|
| Humanity's Last Exam | 9.5% | 38.5% |
| MMMU-Pro (visual reasoning) | 61% | 79.7% |
| Long-context reasoning (AA-LCR) | 45% | 83.3% |
| Terminal-Bench Hard | 12.1% | No data |
| GPQA Diamond | 67.6% | No data |
| IFBench (instruction following) | 67.6% | No data |
| τ²-Bench (agentic tool use) | 36.5% | No data |
| Intelligence Index | No data | 38.1 |
| SciCode | No data | 54.6% |
Independent evaluations by Artificial Analysis, run the same way for every model, at the reasoning setting shown. Percentages unless marked as an index.
Frequently asked questions
Should I choose GPT-5 nano or GPT-6 Luna?
Choose GPT-5 nano for high-volume, cost-sensitive work, writing and code generation and reused prompts and documents (RAG, agents). Choose GPT-6 Luna for very long inputs, reasoning over long documents, the latest release and the most recent knowledge.
Which is cheaper, GPT-5 nano or GPT-6 Luna?
GPT-5 nano is 31% cheaper than GPT-6 Luna at the headline price ($0.1375 / 1M tokens vs $0.20 / 1M tokens).
Which has the larger context window, GPT-5 nano or GPT-6 Luna?
GPT-6 Luna reads more per request: 1.05M tokens against 400k.
What is the difference between GPT-5 nano and GPT-6 Luna?
GPT-5 nano is made by OpenAI and GPT-6 Luna by OpenAI. GPT-5 nano is 31% cheaper than GPT-6 Luna at the headline price ($0.1375 / 1M tokens vs $0.20 / 1M tokens). GPT-6 Luna reads more per request: 1.05M tokens against 400k. GPT-6 Luna is the newer model (Sep 22, 2026, against Aug 7, 2025).
How much do GPT-5 nano and GPT-6 Luna cost per month?
At 20 million input and 5 million output tokens a month, GPT-5 nano costs $3.00 and GPT-6 Luna costs $4.50, at standard rates.
Which is newer, GPT-5 nano or GPT-6 Luna?
GPT-6 Luna was released on Sep 22, 2026 and GPT-5 nano on Aug 7, 2025.