GPT 5.6 Terra

GPT-5.6 Terra is the balanced model in OpenAI's GPT-5.6 series, positioned between flagship Sol and cost-efficient Luna. It is suited for everyday coding, reasoning, agentic work, and general professional tasks.

Reasoning
Vision
Native PDF input
Tool Calling
Structured Output

Added Jul 9, 2026

Pricing

Auto routing · per 1M tokens

Input: $2.00
Output: $12.00
Cache read: $0.20

Compare provider prices

Specifications

Context window: 1.1M
Max output: 128K
Avg output (7d): 365 tokens; Longer than 31% of models

Benchmarks

Artificial Analysis

LMArena

Design Arena

Sourced from Artificial Analysis.

Intelligence Index

42.1

Better than 94% of models compared

Coding Index

76.7

Better than 96% of models compared

Agentic Index

43.2

Better than 84% of models compared

Agentic work

AutomationBench-AA

Workflow automation with guardrail penalties

59.6%

Better than 84% of models compared

Harvey LAB-AA

Legal agentic work criterion pass rate

85.2%

Better than 42% of models compared

AA-Briefcase

Agentic knowledge work (Elo)

1333 Elo

Better than 72% of models compared

GDPval-AA v2

Economically valuable tasks (Elo)

1453 Elo

Better than 80% of models compared

Document reasoning

GDP.pdf

Professional PDF reasoning: all-pass rate

24.0%

Better than 82% of models compared

AA-LCR v1.1

Long context reasoning with updated grading

83.0%

Better than 95% of models compared

MLCR-AA

Medical long-context reasoning

31.7%

Better than 81% of models compared

Reasoning

HLE

Humanity's Last Exam

42.9%

Better than 91% of models compared

IFBench

Instruction-following benchmark

71.2%

Better than 87% of models compared

CritPt

Research-level physics reasoning

30.0%

Coding

Terminal-Bench v4.0

Practical coding and terminal tasks

35.4%

Better than 85% of models compared

SciCode

Python programming for scientific computing

55.0%

Better than 77% of models compared

Knowledge

AA-Omniscience Accuracy

Proportion of correctly answered questions

46.8%

AA-Omniscience Hallucination Rate

Rate of incorrect answers among non-correct responses

87.9%

Legacy benchmarks

GPQA Diamond (legacy)

Graduate-level scientific reasoning

92.5%

Better than 94% of models compared

Terminal-Bench Hard (legacy)

Agentic coding and terminal use

57.6%

Better than 97% of models compared

T²-Bench Telecom (legacy)

Conversational AI agents in dual-control scenarios

86.3%

Better than 79% of models compared

AA-LCR (unversioned / legacy)

Long context reasoning evaluation

83.0%

Better than 95% of models compared

GDPval-AA (unversioned / legacy)

Economically valuable tasks

47.6%

Last updated Oct 14, 2026