GPT-5.6 Terra logo

GPT-5.6 Terra

OpenAI · Released Jun 2026

Conditional

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)

Very Good

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

Very Good

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

Excellent

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

Very Good

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

Good

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

View full pricing

Benchmarks

BenchmarkScoreSource
TerminalBench 2.184.3% Source
Internal CTF (capture-the-flag)91.84% Source

Verdict history

Sep 11, 2026
conditional -> conditional (stale price corrected) — No rating change. The standing summary still priced Terra at $2.50/$15 per 1M six weeks after OpenAI's 30 July cut, while the entity's own pricing fields already read $2.00/$12.00. OpenAI's API pricing page, fetched 11 September 2026, lists gpt-5.6-terra at $2.00 input / $12.00 output short context and $4.00 / $18.00 long context; the API changelog states 'Starting July 30, GPT-5.6 Luna costs 80% less, while GPT-5.6 Terra costs 20% less.' Summary corrected; children that still claimed government-gated preview access or quoted $2.50/$15, Luna at $1/$6 and Sol at $5/$30 were corrected in the same pass.
Jul 22, 2026
pending -> conditional — Reached general availability on July 9, 2026. Pricing held stable at $2.50/$15 through GA. Now accessible to all developers via API and Codex. Remains conditional due to vendor-reported benchmarks and absent ChatGPT integration.
Jul 16, 2026
copy edit to brevity standard, detail preserved in children — Density rewrite: compressed 157w/10s to ~85w/3s per the 55–90w verdict_summary standard. Benchmark scores, pricing tiers, audience fit, and per-task ratings remain in strengths[], benchmarks[], task_strengths[], and verdict_audiences[].

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

How we evaluate