GPT-6.1 Sol and GPT-6 Luna both launched at OpenAI's Sep 29, 2026 DevDay, sharing a 1M-token context and a long-context step at 272K tokens. Luna is the fastest, cheapest GPT-6 tier; Sol sits just below Astra.
GPT-6 Luna is 20× cheaper on input and output and 10× cheaper on cache hits. With a gap that large, the usual answer is Luna for volume and Sol only for requests Luna gets wrong.
| GPT-6.1 Sol | GPT-6 Luna | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Tier | Mid-tier | Budget |
| Input / 1M tokens | $2.00 | $0.10 |
| Output / 1M tokens | $10.00 | $0.50 |
| Cache hit / 1M | $0.10 | $0.01 |
| Cache write / 1M | $2.50 | $0.125 |
| Context window | 1M | 1M |
| Notes | New (Sep 29, 2026 DevDay), below Astra in the GPT-6 family; replaces GPT-6 Sol at the same input/output price with a 50% cheaper cache-read rate; >272K tokens: $4/$15, cache $0.20. ChatGPT Work/Codex on Plus/Pro/Business/Enterprise/Edu, plus API | New (Sep 29, 2026 DevDay), fastest & cheapest GPT-6-family tier; >272K tokens: $0.20/$0.75, cache $0.02 |
GPT-6 Luna is 20× cheaper on input tokens ($0.10 vs $2.00 per 1M). GPT-6 Luna is 20× cheaper on output tokens ($0.50 vs $10.00 per 1M).
| Workload | GPT-6.1 Sol | GPT-6 Luna | Difference |
|---|---|---|---|
| Support chatbot 10,000 conversations × (2,000 input + 400 output tokens) | $80.00 | $4.00 | GPT-6 Luna saves $76.00 |
| Document Q&A 1,000 questions × (30,000 input + 800 output tokens) | $68.00 | $3.40 | GPT-6 Luna saves $64.60 |
| Coding agent session 4M input tokens (90% cache hits) + 150,000 output tokens | $2.66 | $0.15 | GPT-6 Luna saves $2.51 |
Standard rates per 1M tokens from data on this site. The coding-agent row assumes 90% of input is read from the prompt cache (full input price where a model has no cache rate). Cache-write fees, batch discounts and long-context surcharges listed in the notes are not included. Updated November 2, 2026.
requests need more capability than Luna delivers — multi-step coding, complex tool use, nuanced writing — and failures cost more than the 20× token difference.
you are handling high volume — classification, routing, summaries, support chat — where a 20× lower rate dominates any quality gap.
Because Sol costs 20× Luna, a Luna-first cascade stays cheap even with frequent escalation: if 20% of requests go on to Sol, the blend is about 5× Luna — still a quarter of Sol-only. The cheaper Luna pass also works as a filter, for instance deciding whether a support ticket needs Sol at all.
Above 272K input tokens Sol rises to $4/$15 with $0.20 cache hits, Luna to $0.20/$0.75 with $0.02 — the ratio stays at 20× on input and output, so long prompts don't change the decision. At base tiers cache hits are $0.10 vs $0.01 and 30-minute cache writes $2.50 vs $0.125.
OpenAI's Batch API halves both, so Sol in batch is still 10× Luna's standard rate. For offline jobs, try Luna in batch first: it is the cheapest GPT-6-family option available.
20× on input and output ($0.10/$0.50 vs $2.00/$10.00 per 1M tokens) and 10× on cache hits.
Yes, both list 1M tokens, with higher rates above 272K.
GPT-6.1 Sol, at the same input and output price with a 50% cheaper cache-read rate.
How we choose: prices come from each vendor's official pricing page and are re-checked every week. We take no affiliate fees, so rankings aren't for sale.