OpenAI
GPT-5.6 Terra (max)
Released 
July 2026

OpenAI
GPT-5.6 Terra (max)

OpenAI's balanced GPT-5.6 tier for everyday coding, reasoning, and agents at roughly half the cost of Sol.

Modality:
Text
Image
PDF
model ID
openai/gpt-5.6-terra

Output Speed *

119.18
tok/s

Intelligence Index *

42.3
/ 100

Context Window *

1050000
tokens

Input price

12
Anytoken

Output price

72
Anytoken
GPT-5.6 Terra: The Balanced Default for Everyday Agentic and Coding Work GPT-5.6 Terra is OpenAI's mid-tier model in the GPT-5.6 family, sitting between the flagship Sol and the low-cost Luna. It is a reasoning-capable model with a roughly 1.05M-token context window and configurable reasoning effort, positioned as the balanced default for interactive and agentic coding, everyday reasoning, and tool-driven workflows. Terra is competitive with the previous-generation GPT-5.5 while costing materially less, which makes it a practical production workhorse when you need frontier-adjacent capability without paying flagship rates on high-volume traffic. Integrate GPT-5.6 Terra via the AnyAPI.ai API

Performance

Where GPT-5.6 Terra Earns Its Place in the Lineup

Terra's strength is balance: it delivers reasoning and agentic capability close to the prior GPT-5.5 generation while running at a lower tier cost. On Artificial Analysis's Intelligence Index v4.3, GPT-5.6 Terra (max) scores 42, well above the ~24 median for comparable models, and generates output at roughly 112 tokens/second. For production teams, this means Terra can serve as the everyday default across coding, reasoning, and tool-driven tasks without escalating to the flagship Sol tier for most requests. Reserve Sol for the hardest long-running agentic work; route the majority of traffic through Terra.

Benchmarks

GPT-5.6 Terra on Independent Benchmarks

Artificial Analysis places GPT-5.6 Terra (max) at 42 on its Intelligence Index v4.3, well above the median (24) for models in its class, while noting the model is somewhat verbose (around 120M tokens generated during the index). Measured output speed is roughly 112 tokens/second, faster than the ~71 t/s median for reasoning models in a similar price tier. Independent trackers report a high time to first token when reasoning at higher effort settings, which reflects thinking time rather than network latency. Benchmark results depend heavily on the chosen reasoning effort, so validate at your intended setting.

Output Speed

*
119.18
tok/s

Intelligence Index

*
42.3
/ 100

MMLU *

Broad world knowledge and problem-solving
0
%

GPQA *

PhD-level scientific reasoning across physics, biology, chemistry.
93
%

HLE *

Adherence to multi-step structured instructions.
43
%

LiveCodeBench *

Tool-calling reliability in long agentic loops.
0
%

Technical Specifications

What the model supports

GPT-5.6 Terra accepts text and image input and returns text output; audio and video are not supported model modalities. The context window is roughly 1,050,000 tokens with a 128,000-token maximum output. Two specifications drive production decisions: the long-context threshold and reasoning effort. Prompts above 272K input tokens are billed at higher long-context rates across the full request, so very large prompts change your cost model. Reasoning effort (none through max) directly controls latency, token consumption, and quality, making it the primary lever for tuning Terra to a workload.
Verified Specifications — 
GPT-5.6 Terra (max)
*
Input modalities
Text
Image
PDF
output modalities
Text
Context window
1050000
 tokens
Maximum output tokens
128000
Reasoning
Yes
Knowledge cutoff
July 2026
Pricing (standard)
12
 AnyTokens in
 / 
72
 AnyTokens out

Comparison

GPT-5.6 Terra vs GPT-5.6 Sol: When to Escalate

Terra and Sol are siblings in the same GPT-5.6 family, share the naming system and reasoning-effort controls, and both target coding and agentic work. The practical difference is capability ceiling versus cost. Sol is the flagship with the highest reasoning ceiling, built for complex reasoning over large codebases and demanding long-running agentic tasks. Terra is the balanced default that OpenAI positions as competitive with the previous GPT-5.5 generation while costing meaningfully less. For most teams the real decision is not Terra or Sol exclusively, but which tier handles the bulk of traffic and which handles escalations.

Dimension
GPT-5.6 Terra (max)
GPT-5.6 Sol (max)
Context window *
1050000
tokens
1050000
tokens
Output speed *
119.18
tok/s
62.03
tok/s
Intelligence Index *
42.3
47.1
Input pricing
12
AnyToken
12
AnyToken
Output pricing
72
AnyToken
60
AnyToken
Knowledge cutoff *
July 2026
July 2026

Choose GPT-5.6 Terra when you need a cost-efficient everyday default for interactive and agentic coding, general reasoning, and high-volume tool-driven traffic where capability close to the prior generation is sufficient. Choose GPT-5.6 Sol when a task demands the highest reasoning ceiling — complex reasoning across large codebases or long-running multi-agent workstreams — and the added cost is justified by outcome quality. A common architecture runs Terra by default and escalates specific hard cases to Sol.

Limitations & Trade-offs

Where GPT-5.6 Terra (max) falls short

1
Long-context cost cliff: Terra's base pricing is lower than Sol, but requests above 272K input tokens are billed at higher long-context rates (2x input, 1.5x output) across the entire request, not just the tokens over the threshold. For RAG or full-repository prompts that routinely exceed that limit, per-request cost rises sharply. Estimate prompt size, caching behavior, and volume before committing to a large-context architecture, and consider chunking or retrieval trimming to stay under the threshold.
2
Verbosity and reasoning latency: Independent testing notes Terra is somewhat verbose and, at higher reasoning-effort settings, shows a high time to first token driven by thinking time. This inflates output token spend and slows perceived response for interactive UIs. Tune reasoning.effort down for latency-sensitive chat and cap output length; reserve higher effort for tasks where the added quality clearly pays for the extra tokens and delay.
3
Text-and-image input only: Terra accepts text and image input and returns only text. Audio and video are not supported model modalities. Applications needing speech interaction, audio transcription, or media generation must route to specialized models rather than treating Terra as a native multimodal endpoint. This is a meaningful constraint for voice agents or video-analysis pipelines built on a single model.
4
Not the top reasoning ceiling: As the balanced tier, Terra is positioned below flagship Sol on capability. For the hardest problems — complex reasoning over large codebases, demanding long-running agentic workflows — Terra can underperform relative to Sol. If a workload consistently fails at Terra's ceiling, escalating to Sol is the intended path rather than pushing Terra's reasoning effort to max on every request.

Best-Fit Workloads

Where this model earns its place

01

Everyday agentic and interactive coding


OpenAI and GitHub Copilot both position Terra as the balanced default for everyday interactive and agentic coding. Its reasoning-effort control lets you dial capability per task, and its cost tier makes it viable for the high request volumes coding agents generate. Use Terra for code generation, refactoring, and multi-step agent loops, escalating only the hardest reasoning-heavy tasks to Sol.

02

High-volume reasoning at scale


Terra scores 42 on Artificial Analysis's Intelligence Index — well above class median — while sitting in a lower cost tier than Sol. That combination makes it a strong choice for reasoning workloads you run at volume: classification with justification, structured extraction, and analytical assistant turns. Tune reasoning.effort to balance quality against token spend, keeping effort low for latency-sensitive traffic.

03

Tool-driven and function-calling agents


Terra supports function calling via tools and tool_choice plus structured outputs through JSON schema, and independent benchmarks show strong search-and-function-calling scores. This makes it well suited to agents that orchestrate external tools, APIs, and retrieval steps. The large context window helps the agent retain long tool-call histories, though watch the 272K long-context cost threshold on token-heavy sessions.

04

Large-document and codebase analysis


The ~1.05M-token context window lets Terra ingest large repositories, technical documentation, logs, or contracts in a single request. Combined with image input for diagrams and screenshots, it fits document-understanding and code-review pipelines. Note the cost cliff above 272K input tokens: for routinely oversized prompts, retrieval trimming or chunking keeps per-request economics predictable.

Pricing in anytokens via AnyAPI
Input
12
Output
72
Cache write
15
Cache read
1.2

Integration

Access GPT-5.6 Terra (max) via AnyAPI.ai

Access GPT-5.6 Terra (max) through AnyAPI.ai using a unified API built for multi-model AI applications. Integrate GPT-5.6 Terra (max) without maintaining a separate provider-specific connection, and keep the flexibility to test, switch, or combine models as your application requirements evolve.

01

One API integration

Access GPT-5.6 Terra (max) and other AI models through the same API workflow instead of maintaining separate integrations for every provider.

02

Easy model switching

Test GPT-5.6 Terra (max) against alternative models or switch models as your performance, capability, or cost requirements change without rebuilding your application around another provider API.

03

Flexible for production

Use GPT-5.6 Terra (max) from experimentation through production while keeping your AI stack flexible as workloads, traffic, and model requirements evolve.

04

Multi-model applications

Use GPT-5.6 Terra (max) for the workloads where it performs best and combine it with other models for tasks that require different capabilities, performance, or efficiency.

Frequently Asked Questions

Answers to common questions about integrating and using this AI model via AnyAPI.ai

GPT-5.6 Terra is OpenAI's balanced mid-tier model, positioned as the default for everyday interactive and agentic coding, general reasoning, and tool-driven workflows. It offers capability competitive with the prior GPT-5.5 generation at a lower cost tier than flagship Sol, making it a practical production workhorse for high-volume traffic that doesn't require the top reasoning ceiling.

GPT-5.6 Terra has a context window of roughly 1,050,000 tokens and supports up to 128,000 output tokens per request. Note that prompts exceeding 272,000 input tokens are billed at higher long-context rates across the full request, so very large prompts change the cost model.

GPT-5.6 Terra accepts text and image input and returns text output. According to OpenAI's model documentation, audio and video are not supported modalities. Applications needing speech or media generation should route those tasks to specialized models rather than Terra.

Yes. GPT-5.6 Terra is a reasoning-capable model with a configurable reasoning.effort parameter supporting none, low, medium (default), high, xhigh, and max. This lets you trade latency and token cost against answer quality per request — lower effort for interactive UIs, higher effort for harder problems.

Sol is the flagship tier with the highest reasoning ceiling, built for complex reasoning over large codebases and demanding long-running agentic work. Terra is the balanced default at a lower cost tier, competitive with the prior GPT-5.5 generation. Most teams run Terra by default and escalate the hardest tasks to Sol.

* Benchmark data source: Artificial Analysis artificialanalysis.ai