Oleksandr Kuzmenko, EditorPublished 5 min readAI-assisted · editor-reviewed
Why it matters
You can cut agentic computer-use and visual reasoning API expenses by roughly 50% compared to Sol.
TL;DR
- 01GPT-6.1 Terra costs $1/$5 per million tokens, cutting API costs ~50% compared to Sol.
- 02Engineered for desktop automation, computer use, and technical communication at high inference speeds.
- 03Provides a midrange compromise when Luna lacks reasoning depth and Sol is too expensive for high-volume loops.
Key facts
- Input Token Price
- $1.00 / 1M tokens
- Output Token Price
- $5.00 / 1M tokens
- Price Delta vs Sol / GPT-5.6 Terra
- ~50% reduction
- Comparison Tier
- Anthropic Claude Sonnet equivalent
Mid-Tier Positioning and Pricing
OpenAI released GPT-6.1 Terra as a balanced alternative to GPT-6.1 Sol and GPT-6 Luna. It is priced at $1 per million input tokens and $5 per million output tokens, which represents nearly half the operational cost of GPT-5.6 Terra and GPT-6.1 Sol, while remaining roughly 10x the price of lightweight GPT-6 Luna.
Strengths: Computer Use and Technical Communication
The model is tuned specifically for workday computer interactions: navigating user interfaces, processing technical drawings, handling video edits, and drafting proposals. Internal benchmarks reported by OpenAI engineer Ted Sanders indicate that Terra matches Sol and Astra in computer use and visual reasoning at faster inference speeds, while outperforming them in technical communication clarity.
Tradeoffs and API Migration
While Terra excels in execution speed and cost efficiency for desktop agents, it underperforms Sol on deep, unstructured domain reasoning. Developers building agent loops should test Terra on tasks involving high-frequency screen captures and structured tool calls where prompt caching and input token pricing dominate operating budgets.
Try it in 2 minutes
from openai import OpenAI
client = OpenAI()
response = client.chat.completions.create(
model="gpt-6.1-terra",
messages=[
{"role": "user", "content": "Analyze this UI interaction sequence and output next actions."}
]
)python
✓ When to use
- Building UI automation, scraping, or computer-use agents requiring screen evaluation.
- Drafting technical documentation and processing forms where Sol pricing is cost-prohibitive.
- Workflows hitting rate or budget limits on frontier models.
✕ When NOT to use
- Complex multidimensional mathematical reasoning where Sol or Astra remain superior.
- Ultra-cheap bulk text classification where Luna is 10x cheaper.
What to do today
- Benchmark computer-use and screenshot parsing agents against gpt-6.1-terra to test cost savings.
- Review token burn on repetitive Sol workflows to identify candidate pipelines for Terra migration.