Model

GPT-5.6 Terra

CLOSED

OpenAI · GPT-5.6 family · released Jul 9, 2026

Mid tier of the three-tier GPT-5.6 family (GA July 9, 2026 alongside Sol and Luna); OpenAI positions the tiers as 'durable capability tiers that can advance on their own cadence' — Terra is the balanced everyday model between flagship Sol and high-volume Luna. Same 1.05M context, 128K max output and Feb 2026 knowledge cutoff as Sol. Price cut ~20% on July 30, 2026 (launch $2.50/$15 → $2/$12 per MTok), leaving it at 40% of Sol's price. Distinct model tier, not a reasoning-effort configuration of Sol (each tier separately exposes its own effort settings).

ReasoningCodingVisionFunction callingTool useAgentic

2638.6

Elo · rank #27

Parameters

Undisclosed

Active params

—

Context

1.05M tokens

Architecture

Undisclosed architecture (presumed dense transformer); mid 'durable capability tier' of the GPT-5.6 Sol/Terra/Luna family, billed as balancing intelligence and cost

License

Proprietary

Languages

—

API price (in/out)

$2 / $12

Modalities

text · vision

Benchmark results

Bar shows position within the tracked field; marker = field best

Agents' Last ExamAgents50.4%#4

best: GPT-6 Astra · 59.3%

ARC-AGI-1Reasoning96.5%#7

best: Claude Fable 5 · 98.5%

ARC-AGI-2Reasoning83.9%#9

best: GPT-6 Astra · 95.0%

ARC-AGI-3Reasoning0.8%#8

best: GPT-6 Astra · 62.7%

Arena EloHuman preference1466#26

best: Claude Fable 5 · 1504

BrowseCompAgents87.5%#6

best: Kimi K3 · 91.2%

CyberGymCoding81.8%#9

best: MiMo-V2.6-Flash · 95.1%

FrontierBenchAgents20.8%#5

best: Claude Opus 5 · 43.3%

FrontierMath (Tiers 1-3)Math84.9%#2

best: GPT-5.6 Sol · 89.0%

FrontierMath (Tier 4)Math68.3%#2

best: GPT-5.6 Sol · 83.0%

GDPval-AAAgents1593#14

best: Claude Fable 5 · 1932

GPQA DiamondReasoning92.9%#14

best: GPT-6 Astra · 96.0%

Humanity's Last ExamReasoning41.8%#28

best: Claude Opus 5.5 · 67.7%

IFBenchReasoning71.2%#32

best: MiniMax M3 · 83.0%

MMMU-ProVision80.7%#10

best: Claude Opus 4.7 · 85.5%

SWE-bench ProCoding63.4%#10

best: Claude Opus 5.5 · 89.9%

SWE-bench VerifiedCoding95.4%#5

best: Claude Opus 5 · 97.0%

τ³-BankingAgents31.8%#4

best: Step 5 Preview · 42.5%

Terminal-Bench 2.0Coding87.4%#9

best: DeepSeek-V4.1-Flash · 90.6%

Run it locally

Closed weights — available via API only. No local deployment.

Input / M tok

Output / M tok

$12

GPT-5.6 family

Elo progression across releases

GPT-5.6 LunaJul 20262534.3 GPT-5.6 SolJul 20262829.1 GPT-5.6 TerraJul 20262638.6 GPT-5.6 CyberAug 20261000.0

Resources

DOCSdevelopers.openai.com/api/docs/models/gpt-5.6-terra APIAvailable via OpenAI API

API price $2/$12 · each benchmark row carries its own source badge (see methodology)