CrossModel
Back to model catalog

openai · Model guide

GPT-5.6 Terra

openai/gpt-5.6-terra
Modalities
TextImageText
Context
1M
Max output
128K
GPT-5.6 Terra

Balanced tier: close to Sol capability at about half the price

Context window
1.05M
tokens
Max output
128K
tokens
Price vs. Sol
~50%
Same context and modality specs
Sol
Flagship tier, near-frontier reasoning
Terra (this page)
Balanced tier for everyday agentic tasks
Luna
High-throughput tier, cost-first

Terra shares its context window and modality support with Sol, tuned for everyday coding and reasoning.

Overview

GPT-5.6 Terra is the balanced mid-tier model in OpenAI's GPT-5.6 family, sitting between the flagship Sol and the cost-efficient Luna. It launched alongside Sol and Luna — limited preview June 26, 2026, broad API access July 9, 2026 — and targets everyday coding, reasoning, and agentic tasks where capability and cost need to stay in balance.

Terra shares Sol's context window, output limit, and modality support, and runs at roughly half of Sol's per-token price, trading some peak reasoning quality for a meaningfully lower cost per request.

Key capabilities

DimensionDetail
Context window1,050,000 tokens
Max output128,000 tokens
Input modalitiesText, image
Output modalitiesText
Reasoning effortnone, low, medium, high, xhigh, max (default: medium)

Inputs above 272K tokens enter a higher long-context tier (2x input, 1.5x output). Cached input reads get the standard ~90% discount. See live pricing in the model catalog.

Benchmarks

Balanced Tier

Coding ability close to Sol at half the cost

Intelligence Index
55
Sol: 59
Coding Agent Index
77
Sol: 80
Cost per task
$0.55
About half of Sol

Source: Artificial Analysis Intelligence Index and Coding Agent Index.

On Artificial Analysis's Intelligence Index (max reasoning effort), Terra scores 55, four points behind Sol, at roughly half of Sol's per-task cost (about $0.55). On the Coding Agent Index inside Codex, Terra scores 77 — close enough to Sol's 80 for most day-to-day engineering work, while costing noticeably less per run.

When to use it

  • Everyday agentic coding: routine PRs, test writing, and multi-step tool use where Sol-level peak reasoning is not required.
  • High-volume reasoning tasks: workloads where the quality gap versus Sol is small but the request volume is large.
  • Balanced cost/quality default: a sensible default tier before reaching for Sol on the hardest requests or Luna on the simplest ones.

CrossModel exposes GPT-5.6 Terra through an OpenAI-compatible /v1/chat/completions API. Current pricing is available in the model catalog.