gpt-5.6-sol vs gpt-5.6-terra
gpt-5.6-sol vs gpt-5.6-terra: $4/$20 vs $2/$12 per 1M, the same workload priced on both; gpt-5.6-terra is 46% cheaper per request.
| gpt-5.6-sol | gpt-5.6-terra | |
|---|---|---|
| Provider · tier | openai · frontier | openai · mid |
| Input | $4 | $2 |
| Cached input | $0.4 | $0.2 |
| Cache write | $5 | $2.5 |
| Output | $20 | $12 |
| Long-context tier | > 272,000: $8 / $30 | > 272,000: $4 / $18 |
| Per request (2,000 in / 300 out) | $0.014 | $0.0076 |
| Per month at 100k requests | $1,400 | $760 |
| …with the static prefix cached | $860 | $490 |
| Cheaper sibling (cascade) | gpt-5.6-terra | gpt-5.6-luna |
Through TokenZero
Point your SDK at the gateway and keep both keys; the cache formatter, response cache and advisory routing act on every request, and each one comes back with a signed receipt showing the list-price baseline, actual cost and the saving.
OPENAI_BASE_URL=https://tokenzero.sh/w/<key>/v1 ANTHROPIC_BASE_URL=https://tokenzero.sh/w/<key>
Questions
Which is cheaper, gpt-5.6-sol or gpt-5.6-terra?
At list prices (table 2026-10-26) gpt-5.6-terra is 46% cheaper per request on a 2,000-in / 300-out workload. Quality is a separate question: TokenZero's cascade only routes to a sibling it scores as sufficient and measures parity with shadow evaluation.
Where do these prices come from?
The providers' own pricing pages, re-verified on 2026-10-26; the gateway signs receipts with the same table.