Cheaper overall
GPT-6 Luna
99% lower blended rate
Higher benchmark scores
GPT-6 Astra
leads on 7 of 8 figures
Cheaper cached input
GPT-6 Luna
$0.01 vs $1.00, 99% less
Where each one wins
GPT-6 Luna
- Cheaper input tokens $0.10 vs $10.00, 99% less
- Cheaper output tokens $0.50 vs $50.00, 99% less
- Cheaper cached input $0.01 vs $1.00, 99% less
- Cheaper on all four workloads
- Higher long-context reasoning (AA-LCR) 83.3 vs 80.7
GPT-6 Astra
- Higher overall intelligence (Intelligence Index) 52.7 vs 37.3
- Higher expert-exam performance (Humanity's Last Exam) 54.7 vs 38.5
- Higher scientific coding (SciCode) 56.5 vs 54.6
- Ahead on 4 more benchmark figures
What four workloads cost
One run, and the same run a thousand times.
GPT-6 Luna is cheaper on all four workloads, by much the same margin each time — about 100 times, from $0.0004 against $0.0350 on a chat turn to $0.0823 against $8.23 on a long document.
| Workload | GPT-6 Luna | GPT-6 Astra | ×1,000 | Difference |
|---|---|---|---|---|
| Chat turn 1,000 in / 500 out | $0.0004 | $0.0350 | $0.3500 / $35.00 | GPT-6 Luna is 99% cheaper |
| RAG answer 10,000 in / 800 out | $0.0014 | $0.1400 | $1.40 / $140.00 | GPT-6 Luna is 99% cheaper |
| Agent loop 32,000 in / 700 out × 20 calls | $0.0170 | $1.70 | $17.00 / $1,700.00 | GPT-6 Luna is 99% cheaper |
| Long document 400,000 in / 3,000 out | $0.0823 | $8.23 | $82.25 / $8,225.00 | GPT-6 Luna is 99% cheaper |
Price per 1M tokens
Above 272,000 input tokens the rates change, and the new rate applies to the whole request, not just the tokens past that point: GPT-6 Luna's input goes from $0.1000 to $0.2000, and GPT-6 Astra's input goes from $10.00 to $20.00. It does not change which of the two is cheaper.
| Context band | Price | GPT-6 Luna | GPT-6 Astra | Difference |
|---|---|---|---|---|
| Prompts up to 272,000 tokens | Input | $0.10 | $10.00 | GPT-6 Luna is 99% cheaper |
| Prompts up to 272,000 tokens | Cached input | $0.01 | $1.00 | GPT-6 Luna is 99% cheaper |
| Prompts up to 272,000 tokens | Output | $0.50 | $50.00 | GPT-6 Luna is 99% cheaper |
| Prompts over 272,000 tokens | Input | $0.20 | $20.00 | GPT-6 Luna is 99% cheaper |
| Prompts over 272,000 tokens | Cached input | $0.02 | $2.00 | GPT-6 Luna is 99% cheaper |
| Prompts over 272,000 tokens | Output | $0.75 | $75.00 | GPT-6 Luna is 99% cheaper |
Blended: GPT-6 Luna $0.20, GPT-6 Astra $20.00 per 1M — one rate at 3:1 input to output, for comparing two models at a glance.
Both priced by OpenAI. Standard tier, pay-as-you-go.
Other pricing tiers
Both models offer Batch, Flex and Priority, worth up to 50% off their standard rates, and the cheaper model is the same one on every tier — so the tier you buy changes the bill rather than the choice.
Every tier either model sells is here, so this is the whole price sheet. A saving is measured against that model's own standard rate — how we read tiers.
| Tier | GPT-6 Luna | GPT-6 Astra | Cheaper |
|---|---|---|---|
| Batch | $0.05 in / $0.25 out 50% off Standard | $5.00 in / $25.00 out 50% off Standard | GPT-6 Luna |
| Flex | $0.05 in / $0.25 out 50% off Standard | $5.00 in / $25.00 out 50% off Standard | GPT-6 Luna |
| Priority | $0.20 in / $1.00 out | $20.00 in / $100.00 out | GPT-6 Luna |
Benchmarks
They share 8 figures: GPT-6 Luna leads on 1, GPT-6 Astra on 7. The widest gap is 25.4 points, on answer reliability (Omniscience Non-Hallucination).
Overall intelligence
Intelligence Index — Overall intelligence across reasoning, knowledge & math evals
GPT-6 Astra ahead by 15.4
Expert-exam performance
Humanity's Last Exam — Expert-level questions across many academic domains
GPT-6 Astra ahead by 16.2
Long-context reasoning
AA-LCR — Long-context reasoning across large inputs
GPT-6 Luna ahead by 2.6
Scientific coding
SciCode — Research-level scientific coding tasks
GPT-6 Astra ahead by 1.9
Physics research reasoning
CritPt — Unpublished physics research reasoning problems
GPT-6 Astra ahead by 12.3
Real-world task performance
GDPval — Economically valuable, real-world knowledge work
GPT-6 Astra ahead by 8.7
Factual accuracy
Omniscience Accuracy — Breadth of factual knowledge across domains
GPT-6 Astra ahead by 18.8
Answer reliability
Omniscience Non-Hallucination — How reliably the model avoids fabricated answers
GPT-6 Astra ahead by 25.4
Each figure is that model's best published run, with the effort level named beside it — where these scores come from.
Specifications
Both list the same capabilities, so nothing here separates them.
| Specification | GPT-6 Luna | GPT-6 Astra |
|---|---|---|
| Context window | 1,050,000 tokens | 1,050,000 tokens |
| Max output | 128,000 tokens | 128,000 tokens |
| Takes in | Text, image, file | Text, image, file |
| Puts out | Text | Text |
| Knowledge cutoff | 18 May 2026 | 30 Apr 2026 |
| Released | 22 Sep 2026 | 3 Sep 2026 |
| Status | Active | Active |
| Sold by | OpenAI | OpenAI |
| Tool use | Yes | Yes |
| Structured outputs | Yes | Yes |
| Web search | Yes | Yes |
| Prompt caching | Yes | Yes |
| Code execution | Yes | Yes |
| Computer use | Yes | Yes |
| Open weights | Not listed | Not listed |
| Reasoning | Yes | Yes |
FAQs
- Is GPT-6 Luna cheaper than GPT-6 Astra?
- Yes. A 1,000-token prompt with a 500-token reply costs $0.0004 on GPT-6 Luna and $0.0350 on GPT-6 Astra, and the same model is cheaper on every workload on this page.
- How much do GPT-6 Luna and GPT-6 Astra cost per 1M tokens?
- GPT-6 Luna costs $0.10 for input and $0.50 for output. GPT-6 Astra costs $10.00 and $50.00. Standard pay-as-you-go rates, as of 22 Sep 2026.
- Which is better, GPT-6 Luna or GPT-6 Astra?
- On published benchmarks, GPT-6 Astra. It leads on 7 of the 8 figures both models report, including overall intelligence (Intelligence Index), where it scores 52.7 against 37.3.
- Is Batch pricing cheaper for GPT-6 Luna or GPT-6 Astra?
- Batch input costs $0.05 on GPT-6 Luna and $5.00 on GPT-6 Astra, 50% below the standard rate on both.