GPT-6 Sol lists at $2.00/1M / $10.00/1M per 1M tokens. GPT-5.6 Sol lists at $4.00/1M / $20.00/1M. That is about half on stickers, framed as permanent GPT-6 rates, not a launch promo. Same 272K long-context cliff shape. Rates in TokenCalculator checked 23 Sep 2026.
Migrate on one golden prompt
Paste the same production prompt into the cost calculator on GPT-5.6 Sol and GPT-6 Sol. Hold output size fixed, then re-run your eval harness before you cut over.
Rate card: half the sticker
| Meter | GPT-6 Sol | GPT-5.6 Sol |
|---|---|---|
| Input / 1M | $2.00/1M | $4.00/1M |
| Output / 1M | $10.00/1M | $20.00/1M |
| Cache read / 1M | $0.200/1M | $0.400/1M |
| Cache write / 1M | $2.50/1M | $5.00/1M |
| Batch input / output | $1.00/1M / $5.00/1M | $2.00/1M / $10.00/1M |
| Long-context input (>272K) | $4.00/1M | $8.00/1M |
| Long-context output (>272K) | $15.00/1M | $30.00/1M |
| API model id | gpt-6-sol | gpt-5.6-sol |
| Tokenizer in TokenCalculator | Exact (o200k) | Exact (o200k) |
Permanent pricing, not a temporary cut
OpenAI presents GPT-6 Sol standard rates as permanent list prices at roughly half GPT-5.6 Sol. Plan the migration as a durable cut, then confirm quality. Do not treat ChatGPT Work or Codex credits as the same meter as API $/1M.
What stays the same
- About 1.05M context and 128K max output on both catalog rows
- 272K cliff: prompts over 272K input reprice the full request
- Fast mode about 2x standard; Batch about 50% when latency can wait
- Exact o200k counts in TokenCalculator for both
Worked numbers (planning only)
Example A, uncached: 5K input + 1K output. GPT-5.6 Sol about $0.04. GPT-6 Sol about $0.02.
Example B, cache heavy: 100K cached + 5K fresh + 2K output. GPT-5.6 Sol about $0.04 cache + $0.02 fresh + $0.04 output = $0.10. GPT-6 Sol about $0.02 cache + $0.01 fresh + $0.02 output = $0.05.
Migration checklist
- Swap api model id to gpt-6-sol and keep the same system prompt
- Re-run coding and agent evals at equal effort settings
- Price Fast mode and Batch paths separately if you use them
- Watch the 272K cliff on fat RAG agents (rates still double above it)
- Keep GPT-5.6 Sol in the catalog for historical budget math until spend settles
Common mistakes
- Cutting over on sticker savings without equalizing quality
- Assuming ChatGPT credits track the API half-price cut
- Ignoring Fast mode 2x, which can erase half the sticker gain
- Treating curated TokenCalculator rates as a live OpenAI websocket
Frequently asked questions
Is GPT-6 Sol cheaper than GPT-5.6 Sol?
Yes on list rates in this catalog. GPT-6 Sol is $2.00/1M / $10.00/1M. GPT-5.6 Sol is $4.00/1M / $20.00/1M.
Is GPT-6 Sol pricing permanent or promotional?
TokenCalculator catalogs GPT-6 Sol as permanent standard rates at about half GPT-5.6 Sol. Confirm on OpenAI pricing for your account.
Do both models share the 272K cliff?
Yes. Prompts over 272K input tokens reprice the full request at the long-context band on both rows.
Should I migrate from GPT-5.6 Sol now?
If GPT-6 Sol clears your evals, yes for new spend. Price one golden prompt on both, then cut over. Keep 5.6 Sol available until monthly invoices stabilize.
Are token counts Exact for both?
Yes in TokenCalculator. Both use Exact OpenAI o200k encodings.
Does Fast mode still cost more on GPT-6 Sol?
Yes. Fast mode is about 2x standard. Batch remains about half when delayed processing is acceptable.
Where does Luna fit?
GPT-6 Luna lists at $0.100/1M / $0.500/1M. Prefer Luna for high-volume chat when Sol is overkill.
Are these rates live?
No. Curated catalog rates last verified 23 Sep 2026. Confirm on OpenAI pricing.
Next steps
- GPT-6 Sol pricing
- GPT-6 Luna pricing
- Sol vs Opus 5.5
- 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.