GPT-5.6 Terra — Specs, Pricing & Limits
Specs
| Status | Priced |
|---|---|
| Released (API changelog) | 2026-07-09 |
| Context window | 1,050,000 tokens |
| Max output | 128,000 tokens |
| Knowledge cutoff | Feb 16, 2026 |
| Reasoning effort levels | none, low, medium (default), high, xhigh, max |
| Long-context rule | >272,000 input tokens: 2x input, 1.5x output (full request) |
| Cache writes | Billed at the cache-write rate instead of the input rate |
| ChatGPT credit rate (per 1M) | 50 in · 5 cached · 300 out |
Pricing per 1M tokens, every published service tier
| Standard — short context | in $2.00 · cached $0.20 · writes $2.50 · out $12.00 |
|---|---|
| Standard — long context | in $4.00 · cached $0.40 · writes $5.00 · out $18.00 |
| Batch — short context | in $1.00 · cached $0.10 · writes $1.25 · out $6.00 |
| Batch — long context | in $2.00 · cached $0.20 · writes $2.50 · out $9.00 |
| Flex — short context | in $1.00 · cached $0.10 · writes $1.25 · out $6.00 |
| Flex — long context | in $2.00 · cached $0.20 · writes $2.50 · out $9.00 |
| Fast — short context | in $4.00 · cached $0.40 · writes $5.00 · out $24.00 |
| Fast — long context | in $8.00 · cached $0.80 · writes $10.00 · out $36.00 |
| Cache writes, 12-hour lifetime | $5.00 Standard short · $10.00 long — 2.5× the tier's input rate, instead of 1.25× for the default 30-minute lifetime (Responses API) |
Rate limits by usage tier
| Free | Not supported |
|---|---|
| Build | 5,000 RPM · 1,000,000 TPM · 3,000,000 batch queue |
| Launch | 10,000 RPM · 4,000,000 TPM · 200,000,000 batch queue |
| Grow | 15,000 RPM · 40,000,000 TPM · 15,000,000,000 batch queue |
GPT-5.6 Terra was built as the middle of its family — same context window, output ceiling and knowledge cutoff as Sol, at a lower per-token rate — and for a while that made it the obvious model to argue against simply defaulting to Sol. Two price moves have changed what it's competing with.
First, OpenAI cut GPT-5.6 Sol's prices in August, which narrowed the gap between Sol and Terra on every row of the table. Second, GPT-6 Sol arrived in September at Terra's input rate exactly, with an identical cached-input and cache-write rate, a lower output rate and a newer knowledge cutoff. Put the two Terra-priced models side by side and GPT-6 Sol is cheaper on output and newer, on the same window and the same rate-limit group. That doesn't make Terra wrong — a pipeline already tuned on Terra's behavior has a real reason to stay — but it does make "Terra because it's the cheap middle option" an argument that needs re-running against today's table. The third move settles most of it: GPT-6.1 Terra, released October 26, 2026, is the newer Terra on the same window and rate-limit table, cheaper on every Standard price cell, with a deeper cache-read discount and a newer cutoff. OpenAI's GPT-6 guide treats it as this model's successor — see the migration.
Terra shares Sol's rate-limit group rather than sitting in a more restricted one, and its effort levels run from none up to max, with medium as the published default. The long-context rule reprices the entire request once input crosses the threshold, so a workload planned against Terra's lower base rate still has to respect the same cliff.
The table above is computed directly from this site's facts module — every published service tier, the long-context bands, and the rate-limit table.
Verified 2026-10-28 against https://developers.openai.com/api/docs/models/gpt-5.6-terra.
Checked: https://developers.openai.com/api/docs/models/gpt-5.6-terra · https://developers.openai.com/api/docs/pricing