GPT-6 Luna pricing is $0.100/1M input and $0.500/1M output per 1M tokens on the standard tier. Cached input is $0.010/1M. Cache writes are $0.125/1M. Permanent list rates sit well under GPT-5.6 Luna, with a sharper cut on output. Rates in TokenCalculator checked 23 Sep 2026.
Price high-volume chat
Luna is built for volume. Paste a typical turn into the cost calculator, set monthly request count, and compare Haiku 4.5 and GPT-6 Sol on the same text.
GPT-6 Luna rate card
API model id: gpt-6-luna. Context window: about 1.05M tokens. Max output: 128K. Exact OpenAI encodings apply in the TokenCalculator tokenizer.
| Line | USD per 1M tokens |
|---|---|
| Input (<=272K prompt) | $0.100/1M |
| Cached input (<=272K) | $0.010/1M |
| Cache write | $0.125/1M |
| Output (<=272K) | $0.500/1M |
| Batch input | $0.050/1M |
| Batch output | $0.250/1M |
| Long-context input (>272K, whole request) | $0.200/1M |
| Long-context output (>272K, whole request) | $0.750/1M |
| Long-context cached input (>272K) | $0.020/1M |
Why Luna output fell harder than input
Versus GPT-5.6 Luna ($0.200/1M / $1.20/1M), GPT-6 Luna halves input and cuts output from $1.20 to $0.50 per 1M (about 58%). High-volume chat and routing budgets feel that output cut first.
The 272K long-context cliff
Same OpenAI rule as Sol and Astra. Prompts over 272K input reprice the full request at the long-context band. Luna stays cheap, but the cliff still doubles input and raises output. Keep fat RAG dumps under the threshold when you can.
Fast mode and Batch
Fast mode is about 2x standard. Batch is about 50% of standard when overnight latency is fine. Use Batch for offline classification and reranking. Keep interactive chat on standard unless latency SLAs force Fast.
How Luna compares to nearby budget tiers
| Model | Input / 1M | Output / 1M | Cache read / 1M |
|---|---|---|---|
| GPT-6 Luna | $0.100/1M | $0.500/1M | $0.010/1M |
| GPT-5.6 Luna | $0.200/1M | $1.20/1M | $0.020/1M |
| GPT-6 Sol | $2.00/1M | $10.00/1M | $0.200/1M |
| Claude Haiku 4.5 | $1.00/1M | $5.00/1M | $0.100/1M |
Luna undercuts Haiku 4.5 on list rates in this catalog. Haiku may still win if Claude quality or Anthropic tooling is required. Move to Sol when Luna fails coding or multi-step agent evals.
Worked examples
- High-volume turn: 10K fresh input + 2K output about $0.002 on standard rates.
- Same turn with 8K cached: about $0.0013 (cache read $0.01/M).
- Batch halves input and output when overnight latency is fine.
When to use GPT-6 Luna
Use Luna for high-volume chat, routing, light coding, and classification when Sol would overpay. Prefer Sol for harder agent jobs. Prefer Haiku when you standardize on Claude. Prefer Astra or Opus only when quality demands it.
Common mistakes
- Defaulting every agent to Sol or Astra when Luna would clear the eval
- Ignoring the 272K cliff on cheap models (the cliff still doubles input)
- Comparing Luna to Haiku without Approx honesty on Claude counts
- Mixing ChatGPT credits with API Luna rates
Frequently asked questions
How much does GPT-6 Luna cost?
Standard API pricing is $0.100/1M per 1M input and $0.500/1M per 1M output. Cached input is $0.010/1M.
What is the GPT-6 Luna API model id?
Use gpt-6-luna in the OpenAI API.
Is GPT-6 Luna cheaper than GPT-5.6 Luna?
Yes on list rates in this catalog. Input halves. Output falls from $1.20/1M to $0.500/1M.
Does Luna have a long-context surcharge?
Yes. Prompts over 272K input tokens reprice the full request at the long-context band.
GPT-6 Luna vs Claude Haiku 4.5 for high volume?
Luna lists lower on input and output in this catalog. Haiku is Approx in-browser. Price the same prompt on both and pick on quality plus tooling.
When should I use Sol instead of Luna?
When Luna fails coding, long-horizon agents, or hard reasoning evals. Sol is the usual next step before Astra.
Is Fast mode worth it on Luna?
Only when latency SLAs require it. Fast mode is about 2x standard and can erase Luna savings quickly.
Are these rates live?
No. Curated list prices with lastVerified 23 Sep 2026. Confirm on OpenAI pricing.
Next steps
- GPT-6 Sol pricing
- Claude Opus 5.5 pricing
- Cheapest LLM API hub
- OpenAI API pricing hub
- How we verify rates
- Official OpenAI pricing
Related Sep 2026 pricing guides
Frontier rate cards, cost-per-task math, and long-context cliffs. Cross-link these when you compare Sol, Luna, Opus 5.5, Grok 4.7, or Astra.