GPT-5.6 Terra

GPT-5.6 Terra is OpenAI's balanced GPT-5.6 model, generally available July 9, 2026 via the API, Codex, and ChatGPT for Work. It costs $2.50 per million input tokens, $0.25 per million cached input tokens, and $15.00 per million output tokens — exactly half the flagship Sol tier. It ships a 1,050,000-token context window, a 128,000-token maximum output, a February 16, 2026 knowledge cutoff, and text-and-image input with text output. OpenAI positions it as GPT-5.5-competitive at two times lower cost. We tested it hands-on through the API. Score 8.7 out of 10.

Quick Verdict

Score: 8.7 out of 10. GPT-5.6 Terra is the tier most teams should reach for first. It is the middle child of OpenAI's GPT-5.6 family — sitting between the flagship Sol and the economy Luna — and it is priced at $2.50 input and $15.00 output per million tokens, precisely half of Sol. In our own API testing it handled document summarization, structured extraction, a customer-support reply, and ticket routing cleanly on the first attempt, and on one identical cost-math task it produced the same correct answer as the flagship Sol at roughly half the cost and lower latency. It is the right default for high-volume business work. It is the wrong pick if you need frontier reasoning, fine-tuning, or native audio.

• ✅ Matched flagship Sol's answer on our cost-math test at about half the cost and lower latency — the "two times cheaper" thesis held in practice
• ✅ $2.50 input and $15.00 output per million tokens, with a full 1,050,000-token context and no context-length surcharge
• ✅ Nailed four business tasks on the first try, obeying strict output formats such as "JSON only" and "table only"
• ❌ Independent benchmarks put Terra below Sol (Artificial Analysis Intelligence Index 55 versus 59) — it trails the top tier on genuinely hard reasoning
• ❌ No fine-tuning, no native audio, and not selectable in the consumer ChatGPT app

Best for: support teams, operations, and product engineers running high-volume, well-scoped tasks — customer-support automation, document analysis, structured extraction, classification, and internal copilots. Terra is built for companies that send millions of tokens per day and want most of the flagship's quality without the flagship's bill.

GPT-5.6 Terra Pricing at a Glance

All prices are per million tokens and were verified on OpenAI's developer pricing documentation on July 11, 2026. Terra costs exactly half of Sol on every line, and Batch mode halves those figures again. For a plain-English explainer on how input, output, and cached token rates work, see our guide on AI model pricing explained. Sources: OpenAI API pricing and the GPT-5.6 announcement.

How We Tested GPT-5.6 Terra

This is an early hands-on review, not a matured verdict. GPT-5.6 Terra reached general availability on July 9, 2026 — two days before this review — after a gated preview that opened to trusted partners on June 26. We ran the production model, gpt-5.6-terra, directly through the OpenAI Chat Completions API using our own key on July 11, 2026, from Bali. We did not test a demo or a chat wrapper; every result below came back over HTTPS from OpenAI's servers with a 200 status.

Because Terra is pitched as a "balanced everyday model" for high-volume business work, we deliberately built a battery that mirrors that pitch rather than a frontier reasoning gauntlet: an executive-brief summarization of an operations memo, a strict JSON extraction from a messy support email, a customer-support reply draft, and a five-ticket classification-and-routing task. Then, to measure the value claim directly, we sent one identical cost-analysis prompt to both Terra and the flagship Sol and compared the answers, the latency, and the bill. Six calls total, all returned successfully. We report the real outputs and token costs below. Our full methodology mirrors the approach we describe in our explainer on agentic models versus chatbots. Primary references throughout: OpenAI's Terra model card and Artificial Analysis for independent scores.

What Is GPT-5.6 Terra?

GPT-5.6 Terra is the balanced member of OpenAI's GPT-5.6 family, released alongside the flagship Sol and the economy Luna on July 9, 2026. In OpenAI's new naming system the number denotes the generation while Sol, Terra, and Luna are durable capability tiers rather than model sizes — the same idea persists across releases, so "Terra" should keep meaning "the balanced tier" in future generations. Terra's model ID is gpt-5.6-terra, and it targets the workloads that dominate real business volume: customer support, internal tools, and document analysis.

OpenAI's headline claim for Terra is specific and quotable: "GPT-5.5-competitive performance at two times lower cost." In other words, Terra aims to land near the quality of last quarter's flagship, GPT-5.5, while charging half as much per token. GPT-5.5 has not been deprecated; it remains available, which makes Terra a genuine price-for-quality alternative rather than a forced migration. Terra shares the family's full plumbing — a 1,050,000-token context window, a 128,000-token maximum output, and a February 16, 2026 knowledge cutoff — so choosing the cheaper tier does not cost you context. Where Terra steps back from Sol is raw capability and the new multi-agent "ultra" reasoning mode, which OpenAI concentrates on the flagship. Details are documented on OpenAI's models reference and summarized in the launch post.

Key Features

1,050,000-token context window

Terra carries the same 1.05M-token context window and 128,000-token maximum output as the flagship Sol. That parity matters: many vendors reserve the biggest context for the top tier, so getting a full million-token window on the balanced model removes a common reason to pay up. Terra reads long contracts, multi-document bundles, and large codebases without fragmenting the input across calls. Pricing is flat with no context-length surcharge, which is a meaningful contrast with models that raise rates past a token threshold. See OpenAI's Terra model card and its pricing docs.

Reasoning effort scale from low to max

Every Terra API call accepts a reasoning-effort parameter that ranges from low through xhigh up to the new max level. Reasoning tokens are billed at the output rate and share the completion-token budget, so a high-effort run on a dense problem spends more and can truncate the visible answer if your token cap is set too low — we saw Terra emit reasoning tokens even on a low-effort classification job. The multi-agent "ultra" tier that orchestrates several sub-agents is a flagship Sol feature; on Terra you have the low-to-max ladder, which is plenty for the high-volume tasks it targets. Documented on the OpenAI models reference; behavior confirmed in our own Artificial Analysis-aligned testing.

Programmatic Tool Calling and the agentic stack

Terra includes Programmatic Tool Calling, where the model writes and runs JavaScript in an isolated, ephemeral runtime to orchestrate tool calls in code rather than one call at a time — and it is compatible with Zero Data Retention. On top of that it ships the full tool suite: function calling, structured outputs, streaming, web search, file search, image generation, code interpreter, hosted shell, computer use, and MCP (Model Context Protocol) clients. For a balanced tier to inherit the complete agentic toolbox, rather than a stripped-down subset, is unusual and genuinely useful. References: Terra model card, GPT-5.6 launch.

Improved prompt caching and Batch API

Cached input on Terra is billed at $0.25 per million tokens — a 90 percent discount on the $2.50 standard input rate — with cache writes at 1.25 times input and a 30-minute minimum cache lifetime. For support bots and copilots that re-send a large system prompt on every turn, caching is the single biggest lever on effective cost. Terra also supports the Batch API at half price ($1.25 input and $7.50 output per million tokens) for non-urgent jobs like overnight document processing. Verified against OpenAI's pricing documentation and the announcement.

Text and image input, text output

Terra accepts text and image input and returns text. There is no native audio input or output, and image generation is a callable tool rather than an output modality — the same shape as the rest of the GPT-5.6 family. For document-heavy business work, which is Terra's lane, vision input covers scanned invoices, charts, and screenshots, and the text-only output keeps responses predictable for downstream parsing. Anyone who needs a cheaper multimodal option should also weigh Gemini 3 Flash. Sourced from the Terra model card and models reference.

No fine-tuning at launch

Fine-tuning is not supported on Terra at launch. Teams that depend on tuned production variants cannot bring them to Terra yet and will need to stay on an older fine-tunable model. This is the same gap the flagship Sol has, and OpenAI has not published a timeline. If tuned behavior is core to your stack, factor this in before migrating. Confirmed on the Terra model card; see also the pricing docs for the full family.

GPT-5.6 Terra Pricing in 2026

Terra's pricing is deliberately simple: one flat per-token rate with no context tiers, plus the usual Batch and Priority modes. Every figure below was verified on OpenAI's developer pricing page on July 11, 2026.

API pricing (per million tokens)

There is no long-context surcharge on Terra — the rate is the same whether your prompt is a thousand tokens or a million. That flat structure is easier to forecast than tiered pricing, and it is one of the reasons Terra is attractive for finance and operations teams that need predictable unit economics. The most important comparison is internal: Terra at $2.50 and $15.00 is exactly half of GPT-5.5 at $5.00 and $30.00, and OpenAI's pitch is that Terra stays competitive with GPT-5.5 at that halved rate. It also matches what the previous mid-tier GPT-5.4 charged at standard rates, so long-time OpenAI customers get a familiar price with a newer base. References: OpenAI pricing, GPT-5.6 announcement.

Best for: teams running steady, high-volume traffic — support queues, extraction pipelines, and internal assistants — where halving the per-token rate against GPT-5.5 or Sol turns directly into margin. Solo builders and prototypes should test Luna first, then step up to Terra when quality demands it.

What We Found Testing Terra

Here is what actually came back from gpt-5.6-terra over the API. All six calls succeeded; latencies ran between 1.8 and 3.4 seconds on our small prompts, and the four business tasks together cost about one cent in tokens.

Document analysis and summarization

We handed Terra an internal operations memo and asked for an executive brief: three headline metrics, the top risk, and one recommendation, under 130 words. It returned a tight, correctly structured brief — ticket volume up 18 percent quarter-over-quarter to 41,320, median first response improved from four hours twelve to two hours forty-seven, CSAT held at 91 percent — and it correctly derived the 34 percent improvement in response time that was not stated explicitly in the source. In our testing, Terra respected the word budget and the requested structure without hand-holding. This is exactly the kind of routine synthesis that used to justify a flagship, and Terra did it in 3.4 seconds for a fraction of a cent.

Structured extraction into JSON

We gave Terra a messy customer email and a fixed schema, instructing it to output only valid minified JSON. It returned clean, parseable JSON on the first try — order ID, issue types drawn from the allowed set, requested actions, sentiment "negative," churn risk "high," the phone number, and the loyalty signal — with no prose wrapper and no stray Markdown fences. For a support or operations pipeline that feeds a database, that reliability on "output only JSON" is worth more than a benchmark point. It is the single most common production task we throw at a mid-tier model, and Terra passed cleanly.

Customer-support reply

Asked to draft a warm, specific support reply under 120 words — apologize for two missed emails, confirm the correct item will ship, refund a $12 fee, give a realistic timeframe, and retain the customer — Terra produced a reply we would send with light editing. It apologized without groveling, committed to a one-to-two business day processing window with a five-to-eight day delivery estimate, and acknowledged the customer's third order. It used bracketed placeholders for the names, which is the correct behavior for a template. Honest note: the timeframe it invented is plausible but unverified, so a human still owns the final promise — as it should for any customer-facing message.

Classification and routing

We sent five support tickets and asked for a compact Markdown table with category, priority, and route. Terra returned exactly that — a double-charge billing issue as P1 to Finance, an API 500 error as P1 Bug to Engineering, a how-to as P3 to customer support, an annual-billing discount request as P2 to Sales, and a password-reset failure as P2 Auth to support. The routing logic was sensible and the output was strictly the table we asked for, nothing else. One detail worth flagging: even at low reasoning effort, Terra spent a handful of reasoning tokens on this task, which bill at the output rate — a small but real line item at scale.

The value test: Terra versus Sol on the same task

The most revealing test was the simplest. We sent one identical cost-analysis prompt — compute the daily and 30-day spend for a workload of eight million input and two million output tokens per day on two price schedules, then recommend when the pricier model is justified — to both Terra and the flagship Sol. Both returned the same correct numbers: $50 per day and $1,500 per month on the cheaper schedule, $100 per day and $3,000 per month on the pricier one, a $1,500 monthly saving, and the sensible recommendation to default to the cheaper model unless the pricier one adds more than $1,500 of value. The difference was in the receipt. Terra answered in 2.4 seconds and cost about $0.0044 in tokens; Sol answered in 3.2 seconds and cost about $0.0086 — nearly twice as much for an equivalent answer. On a well-scoped business task, Terra matched the flagship at half the price and finished faster.

The honest caveat: we chose a task Terra handles well. On genuinely hard reasoning, independent scores say Sol should pull ahead — Terra sits four points below Sol on the Artificial Analysis Intelligence Index. Our battery validates the value pitch for routine, high-volume work; it does not claim Terra equals the flagship on frontier problems, and we would not make that claim from an early hands-on.

Benchmarks: How Terra Ranks

Independent numbers matter more than vendor slides, so we separate the two. The clearest third-party read comes from Artificial Analysis, which places Terra squarely in the middle of the GPT-5.6 family.

Read of the table: on the independent Artificial Analysis Intelligence Index, Terra scores 55 against Sol's 59 and Luna's 51 — a real four-point gap below the flagship, but comfortably above the economy tier. On the Artificial Analysis Coding Agent Index it scores 77 against Sol's 80. According to OpenAI's own figures, Terra hits 87.4 percent on Terminal-Bench 2.1, close behind Sol's 88.8 percent; we label that self-reported because OpenAI published it and it has not been independently reproduced. One honest gap: Terra was not submitted to the independent SWE-bench Verified leaderboard, so there is no third-party agentic-coding score to cite yet — the Coding Agent Index of 77 is the closest public reference. We do not cite GPQA Diamond, AIME, or MMLU figures for Terra because no reliable standalone results exist. Sources: Artificial Analysis and the OpenAI GPT-5.6 announcement.

Terra vs Sol vs Luna, and vs GPT-5.5

The most useful comparison for Terra is inside its own family. All three tiers share the plumbing — a 1.05M-token context, the same tool stack, the same February 16, 2026 cutoff — and differ on capability and price.

Pick Terra for the broad middle: support automation, document analysis, extraction, and internal tools where you want most of the flagship's quality at half its price. Step up to Sol when a task is genuinely hard — complex agentic coding, long-horizon planning, or work that benefits from the multi-agent ultra reasoning tier — and the four-point Intelligence Index gap will actually change the outcome. Drop to Luna for the highest-volume, lowest-stakes jobs — bulk summarization, drafting, and routine automation — where speed and cost beat marginal quality; if you are weighing Luna, also look at Gemini 3 Flash.

Against the outgoing flagship GPT-5.5, Terra is the value play OpenAI designed it to be: the same $2.50 and $15.00 rates that make it half of Sol also make it half of GPT-5.5, and OpenAI claims competitive quality at that price. If your workload runs on GPT-5.5 today and does not lean on fine-tuning, Terra is the obvious A/B test. Teams comparing across vendors should also benchmark Claude Sonnet 5, Anthropic's balanced tier, and can use Claude Fable 5 or Grok 4.3 as frontier reference points. As always, run your own LLM evaluation on your real prompts before committing. Independent scores via Artificial Analysis; pricing via OpenAI.

Pros and Cons After Testing

What we liked

• The value thesis held under test. On an identical task, Terra matched flagship Sol's answer at roughly half the token cost and lower latency — exactly what "GPT-5.5-competitive at two times lower cost" is supposed to mean.
• Disciplined instruction-following. It obeyed "output only JSON" and "table only" cleanly, respected word budgets, and did not pad — the behaviors that make a mid-tier model safe to wire into a pipeline.
• Full context, no penalty. A 1,050,000-token window and 128,000-token output identical to the flagship, with flat pricing and no context-length surcharge.
• Fast on business-sized prompts. Responses landed in 1.8 to 3.4 seconds in our runs, snappy enough for interactive support and internal copilots.
• Effective cost is even lower than the sticker. Cached input at $0.25 per million tokens (a 90 percent discount) and a Batch API at half price push real-world spend well below the headline rate.
• The complete agentic toolbox. Programmatic Tool Calling, function calling, structured outputs, web and file search, code interpreter, computer use, and MCP all ship on the balanced tier, not just the flagship.

Where it falls short

• It is not the flagship. Terra sits four points below Sol on the independent Artificial Analysis Intelligence Index (55 versus 59); on genuinely hard reasoning, Sol should win.
• No fine-tuning at launch. Teams that rely on tuned production variants cannot migrate them to Terra yet, and there is no announced timeline.
• Invisible in consumer ChatGPT. Terra is available through the API, Codex, and ChatGPT for Business and Enterprise only — individual ChatGPT users get Sol, not Terra.
• Text and image in, text out only. No native audio and no image generation as an output; image creation is an external tool call.
• A benchmark data gap. Terra was not submitted to SWE-bench Verified, so there is no independent agentic-coding number at launch — only the Coding Agent Index of 77.

Real-World Use Cases

High-volume customer support automation

Terra's clean instruction-following and structured outputs make it a strong Tier-1 support engine — drafting replies, extracting order details, and routing tickets. In our tests it produced a sendable support reply and a correct routing table on the first try, and cached input pricing keeps the recurring system-prompt cost low at volume.

Document analysis and summarization at scale

With a 1.05M-token context and reliable summarization, Terra is well suited to condensing contracts, reports, and research bundles into briefs. Our operations-memo test showed it deriving metrics that were implied but not stated, which is the useful part of summarization rather than mere extraction.

Structured data extraction

Feeding unstructured text — emails, forms, transcripts — into a fixed JSON schema is Terra's sweet spot. It returned valid, parseable JSON with no wrapper in our extraction test, which is exactly what a database-backed pipeline needs.

Ticket classification and triage

Routing queues by category, priority, and destination is a natural fit. Terra's routing logic was sensible and its output format was strict, so the results drop straight into a workflow tool without post-processing.

Internal copilots and business tools

For internal assistants that answer employee questions or draft first-pass content, Terra offers most of the flagship's quality at half the cost — the right trade for tools used all day by many people.

Long-context RAG on a budget

The full million-token window plus 90-percent-off cached input make Terra viable for retrieval-augmented pipelines over large corpora without paying flagship rates or hitting a context surcharge.

Batch and back-office pipelines

For non-urgent overnight jobs — bulk classification, translation, or document processing — the Batch API halves Terra's rate to $1.25 input and $7.50 output per million tokens, making large runs affordable.

Programmatic tool-calling agents

Because Terra can write and run JavaScript to orchestrate tool calls, it suits agents that stitch several APIs together in one turn — data cleanup, enrichment, and transformation flows — without a hand-written function-call harness.

Frequently Asked Questions

What is GPT-5.6 Terra?

GPT-5.6 Terra is the balanced capability tier of OpenAI's GPT-5.6 family, generally available July 9, 2026 via the API, Codex, and ChatGPT for Work. It sits between the flagship Sol and the economy Luna, carries a 1,050,000-token context window and a 128,000-token maximum output, and accepts text and image input with text output. OpenAI positions it as GPT-5.5-competitive at two times lower cost. Its model ID is gpt-5.6-terra.

How much does GPT-5.6 Terra cost?

GPT-5.6 Terra costs $2.50 per million input tokens, $0.25 per million cached input tokens, and $15.00 per million output tokens — exactly half of the flagship Sol. Batch mode halves those rates to $1.25 input and $7.50 output, and Priority processing doubles them to $5.00 and $30.00. There is no context-length surcharge. Prices were verified on OpenAI's developer pricing page on July 11, 2026.

How is GPT-5.6 Terra different from GPT-5.6 Sol?

Terra is the balanced tier and Sol is the flagship. They share the same 1,050,000-token context, the same tool stack, and the same February 16, 2026 knowledge cutoff, but Sol is more capable and costs twice as much ($5.00 input and $30.00 output per million tokens). On the independent Artificial Analysis Intelligence Index, Sol scores 59 to Terra's 55. The multi-agent ultra reasoning tier is a Sol feature. In our testing, Terra matched Sol's answer on a well-scoped cost-math task at half the price.

Is GPT-5.6 Terra better than GPT-5.5?

OpenAI positions Terra as competitive with GPT-5.5 while costing half as much, at $2.50 input and $15.00 output per million tokens versus GPT-5.5's $5.00 and $30.00. GPT-5.5 has not been deprecated and remains available. Independent benchmarks for Terra are still limited, so the honest answer in July 2026 is that Terra is designed to deliver similar quality at half the price for most business tasks — worth an A/B test on your own prompts, especially if you do not rely on fine-tuning.

What is GPT-5.6 Terra's context window?

GPT-5.6 Terra has a 1,050,000-token context window and a maximum output of 128,000 tokens — identical to the flagship Sol. Pricing is flat with no long-context surcharge, so the per-token rate is the same whether your prompt is a thousand tokens or a million. Reasoning tokens count against both the context window and the output billing rate.

Can you use GPT-5.6 Terra in ChatGPT?

Not in the consumer ChatGPT app. Terra is available through the OpenAI API, inside Codex, and in ChatGPT for Business and Enterprise (Work) plans. Individual ChatGPT Plus and Pro users can select the flagship Sol but not Terra or the economy Luna, which are aimed at developers and organizations.

Does GPT-5.6 Terra support fine-tuning?

No. Fine-tuning is not supported on GPT-5.6 Terra at launch, the same as the flagship Sol. Teams that run tuned production variants cannot migrate them to Terra yet and should stay on an older fine-tunable model until OpenAI adds support. No timeline has been announced.

Is GPT-5.6 Terra multimodal?

Partly. GPT-5.6 Terra accepts text and image input and produces text output. It does not support native audio input or output, and it does not generate images as an output modality — image generation is a callable tool. For document work, vision input covers scanned pages, charts, and screenshots, which is what Terra's target audience needs.

What did GPT-5.6 Terra score on independent benchmarks?

On the independent Artificial Analysis Intelligence Index, Terra scores 55, versus 59 for Sol and 51 for Luna. On the Artificial Analysis Coding Agent Index it scores 77, versus 80 for Sol. OpenAI self-reports 87.4 percent on Terminal-Bench 2.1. Terra was not submitted to the independent SWE-bench Verified leaderboard, so there is no third-party agentic-coding score yet, and there are no reliable standalone GPQA, AIME, or MMLU figures for it.

What is GPT-5.6 Terra best for?

Terra is best for high-volume, well-scoped business work: customer-support automation, document analysis and summarization, structured data extraction, ticket classification, internal copilots, and long-context retrieval on a budget. It is built for teams that process millions of tokens per day and want most of the flagship's quality at half the cost. It is not the right pick for frontier reasoning, fine-tuning, or native audio.

Does GPT-5.6 Terra support function calling and structured outputs?

Yes. Terra supports function calling, structured outputs, streaming, and Programmatic Tool Calling, where the model writes and runs JavaScript in an isolated runtime to orchestrate tool calls. It also has native web search, file search, code interpreter, computer use, and MCP client support. In our testing it reliably returned strict JSON and a clean Markdown table when asked, which is what structured-output workflows depend on.

Should I choose GPT-5.6 Terra or Luna?

Choose Terra when quality matters and you want most of the flagship's capability at a mid-tier price — it scores 55 on the Artificial Analysis Intelligence Index versus Luna's 51. Choose Luna, at $1.00 input and $6.00 output per million tokens, for the highest-volume, lowest-stakes work where speed and cost outweigh marginal quality, such as bulk drafting and routine automation. A common pattern is Luna for simple traffic and Terra for anything that needs more judgment.

Verdict: 8.7 out of 10

GPT-5.6 Terra earns an 8.7 out of 10 for one clear reason: it delivers most of a flagship's usefulness at half the price, and in our hands-on testing that trade held up. It matched flagship Sol's answer on an identical task at roughly half the token cost and lower latency, obeyed strict output formats across four business tasks, and ships the full million-token context and complete agentic toolbox rather than a stripped-down subset. What keeps it from a higher score is honest and specific: it sits four points below Sol on the independent Artificial Analysis Intelligence Index, it has no fine-tuning at launch, it is invisible in the consumer ChatGPT app, and it was never submitted to SWE-bench Verified, leaving a real data gap under a business-first pitch.

Score breakdown:

• Features: 8.5 out of 10 — full 1.05M context, the complete agentic tool stack, Programmatic Tool Calling, and reasoning effort up to max. Held back by no fine-tuning, no native audio, and the missing multi-agent ultra tier reserved for Sol.
• Ease of Use: 8.9 out of 10 — a drop-in gpt-5.6-terra model ID on the standard Chat Completions and Responses APIs, flat pricing with no context surcharge, and clean, predictable structured outputs in our tests.
• Value: 9.2 out of 10 — the strongest column. Half the price of GPT-5.5 and Sol for competitive quality on routine work, cached input at a 90 percent discount, and Batch at half again. Our value test showed it matching the flagship at half the cost.
• Support: 8.3 out of 10 — OpenAI's developer docs, model cards, and pricing pages are clear and current, but Terra is API, Codex, and Work only, so there is no consumer-app on-ramp and it is only two days into general availability.

Final word: Choose GPT-5.6 Terra as your default tier if you run high-volume, well-scoped business work and want to halve your model bill without gutting quality — support automation, extraction, summarization, and internal tools are exactly its lane. Step up to the flagship Sol only when a task is genuinely hard enough that a four-point Intelligence Index gap changes the outcome, and drop to Luna for the cheapest, highest-volume traffic. For most teams, most of the time, Terra is the correct first choice — run it against your own prompts alongside Claude Sonnet 5 and decide from your own numbers. Last tested: July 11, 2026.

Sources

• OpenAI — Introducing GPT-5.6 (Sol, Terra, Luna)
• OpenAI — GPT-5.6 Terra model card
• OpenAI — API pricing
• Artificial Analysis — independent LLM benchmarks (Intelligence Index, Coding Agent Index)
• ThePlanetTools.ai hands-on API testing of gpt-5.6-terra, July 11, 2026 (six calls via the Chat Completions API).