01Overview02Models03Rankings04Compare05Quiz06Guide

Model

GPT-5.6 Terra NO.08

The GPT-5.6 balanced workhorse — GPT-5.5-class quality at half the price.

CompareBack to the index

OpenAI · GPT · Balanced · Closed weights

Specification

Fig.01 — GPT — OpenAI

USA

Context window
1.1Mtokens
Max output
128Ktokens
Input price
$2.50/ 1M
Output price
$15/ 1M
Cached input
$0.25/ 1M
Throughput
141tok/s
Class
Balanced
Modalities
Text, Image, Pdf
Weights
Closed
Knowledge cutoff
Feb 2026
Released
Jul 2026

Overview

GPT-5.6 Terra is the balanced, everyday-work member of the GPT-5.6 family, tuned for high-volume business tasks like customer support, internal tools, and document analysis. OpenAI positions it as matching the previous GPT-5.5 flagship at half the cost, and on Artificial Analysis' Coding Agent Index (77.4) it lands just above Claude Fable 5 — in roughly a third of the time and at about a quarter of the estimated cost. It shares the family's ~1.05M-token context, 128K output, and February-2026 knowledge, streams quickly (~141 tok/s), and is the sensible default for most production workloads that don't need Sol.

Strengths

  • GPT-5.5-class quality at half the price
  • Strong agentic-coding value (Coding Agent Index 77.4)
  • Fast output (~141 tok/s) for a balanced model
  • ~1.05M-token context with full multimodal input
  • Programmatic tool calling and predictable caching

Trade-offs

  • Below Sol on the hardest long-horizon work
  • Pricier than Luna for high-volume bulk jobs
  • No native audio or video input

Fit

Best for

  • Everyday production assistants and agents
  • Customer support and internal tools at volume
  • Document analysis and summarization
  • Cost-aware coding and refactoring

Not ideal for

  • The very hardest autonomous coding (use Sol)
  • Ultra-cheap high-volume classification (use Luna or nano)

Capabilities

Ten axes, normalized 0–100 and scored the same way across the whole catalog — so a 78 here means what a 78 means anywhere else on the bench.

Fig.02 — Capability radar

0 — 100

Reasoning
91
Coding
91
Math
90
Writing
89
Knowledge
90
Speed
74
Agentic
90
Vision
87
Multilingual
89
Long Context
90

Benchmarks

Public results, with independent third-party runs marked. Bars normalize percentages against 100 and Elo ratings against a 1500 ceiling.

Independent index

Artificial Analysis Intelligence Index

Composite of ~9–10 independent evals · Artificial Analysis

55/ 100

BenchmarkResult
Artificial Analysis Intelligence IndexIndependentGeneral · Artificial Analysis
55%
SWE-bench ProIndependentCoding · SWE-bench Pro
63.4%

Alternatives

Models in roughly the same class — the ones worth weighing against this record.

Compare side by side
01GPT-5.6 SolOpenAIOpenAI's GPT-5.6 flagship — its best coding-and-agentic model yet, for the hardest work.02Claude Sonnet 5AnthropicThe most agentic Sonnet yet — near-Opus coding reliability at Sonnet cost.03GPT-5.5OpenAIPrevious-generation OpenAI flagship — frontier-reliable, now succeeded by GPT-5.6 Sol.

Index

PreviousGPT-5.6 SolNextGPT-5.6 Luna
All models/EDUCATION