GPT-5.6 Terra
OpenAI · Released Jun 2026
GPT-5.6 Terra is OpenAI's mid-tier workhorse: GPT-5.5-class capability on OpenAI's own claim, at $2.00 input and $12.00 output per 1M tokens at short context ($4.00 / $18.00 long) after the 20% price cut of 30 July 2026. It is generally available via the API and Codex. Remains conditional: all benchmarks are vendor-reported without independent reproduction, and Terra is not selectable in the standard ChatGPT model picker.
Is it right for you?
Good for
- Cost-efficiency sweet spot: GPT-5.5-class capability on OpenAI's own claim, at $2.00 input / $12.00 output per 1M short context. Terra is priced to replace GPT-5.5 as the default model for most production workloads, if the quality claim holds.
- TerminalBench 2.1 at 84.3% ties Claude Fable 5, Anthropic's publicly available flagship, at the same score. For agentic coding that doesn't need Sol's extreme reasoning depth, Terra is the rational pick.
- High-volume enterprise workloads: customer support, internal tools, document analysis, and production pipelines where reliability matters but per-request cost needs to scale.
- Strong cybersecurity capability for its tier: scored 91.84% on internal CTF testing. Carries activation classifiers and the same safety stack as Sol for regulated cyber and bio workflows.
Not good for
- Long-context jobs budgeted off the headline rate: past short context Terra bills $4.00 input and $18.00 output per 1M tokens.
- No ultra or max reasoning modes: Terra lacks the subagent decomposition and extended deliberation that distinguish Sol. For the hardest agentic work, you still need Sol.
- The 84.3% TerminalBench score is only 0.9 points ahead of GPT-5.5 at 83.4%. The 'competitive with GPT-5.5' claim is directionally right but the gap is small: your own evals will tell you more than the scoreboard.
How it performs by task
Agentic coding (mid-tier)
84.3% on TerminalBench 2.1 ties Claude Fable 5. Strong enough for most production coding agents. Sol is the upgrade for the hardest problems.
Cybersecurity
CTF score of 91.84%. Robust security capabilities with activation classifiers. Good for defensive workflows and vulnerability analysis without Sol's premium.
Cost-efficient production workloads
At $2.00/$12.00 per 1M short context, Terra is the balanced cost-to-capability pick in the GPT-5.6 family, above Luna's price floor and well under Sol. OpenAI's claim of GPT-5.5-class performance is the most commercially important headline for most teams.
General reasoning and everyday work
Pitched as the balanced everyday model. For routine reasoning, writing, analysis, and business intelligence, Terra should be indistinguishable from Sol in quality.
Deep research and hard reasoning
Lacks Sol's max reasoning and ultra subagent modes. Falls behind Sol, Mythos 5, and Fable 5 on the hardest problems. Route hard tasks to Sol, not Terra.
Pricing
Input
$2.00 / 1M tokens
Output
$12.00 / 1M tokens
Context
Short $2.00/$12.00 · long $4.00/$18.00 per 1M
Benchmarks
Verdict history
Sources
Verification log
- Pricing— Updated
Automated agent
Rates re-verified: pricing page row gpt-5.6-terra, verbatim 'Short context input $2.00 | Short context cached input $0.20 | Short context output $12.00 | Long context input $4.00 | Long context cached input $0.40 | Long context output $18.00'. API changelog Jul 30 2026: 'GPT-5.6 Terra costs 20% less'. Stored pricing_input/output ($2.00/$12.00) already correct; the prose was not (same stale self-authored prose class as gpt-5-6-luna, found while verifying Luna). FIXED: pricing_context 'Mid-tier. 20% cut Jul 30 -> $2.00/$12.00.' -> 'Short $2.00/$12.00 · long $4.00/$18.00 per 1M'; pricing_url openai.com Sol preview blog -> developers pricing page; removed false avoid_for 'Government-gated limited preview: ... only ~20 trusted partners today' (contradicted by its own 2026-07-22 GA transition); removed false not_for_whom 'Anyone without trusted-partner access ... Terra is gated'; corrected '$2.50/$15' in a strength and a task_strength, and 'Luna ($1/$6) ... Sol ($5/$30)' in an audience to $0.20/$1.20 and $4/$20; dropped the now-unsupported 'half the cost/price' ratio; reworded the task_strength that called Terra 'the best cost-to-capability ratio in the family', which contradicted the Luna entity's own claim; added a long-context rate avoid_for. Summary still said '$2.50/$15': corrected via standalone verdict_set (conditional -> conditional, verdict 8f5982bc-9cdc-4539-8fa6-5fec4135dde7), no announcing post (six-week-old price cut, not news), no draft.
- Pricing— No changes
Automated agent
- Pricing— No changes
Automated agent
- Pricing— No changes
Automated agent
- Pricing— No changes
Automated agent
- Status— Updated
Automated agent
Summary re-dated to Jul 7 2026 and made explicit about what is awaited: stays pending pending GA + independent benchmarks; still government-gated (~20 orgs); family trending toward release (Sol Ultra confirmed for Codex Jul 6, ~Jul 9 GA per prediction markets, unconfirmed). Verdict unchanged (pending).
- Pricing— No changes
Automated agent
- Pricing— No changes
Automated agent
Pricing unchanged: $2.50/$15 per 1M tokens. Government-gated. Confirmed by OpenAI blog (GPT-5.5-class at half cost).
- Pricing— No changes
Imported at launch
- Profile— No changes
Imported at launch