OpenAI · Performance on a more reasonable budget
GPT-5.6 Terra
A practical compromise for solving complex problems and autonomous programming. Slightly less capable than the Sol variant, but more economical.
- Costs
- Upper mid-range · from July 30, 2026, standard pricing is $2 per million input tokens and $12 per million output tokens
- Speed
- Medium depending on processing thoroughness
- Availability
- Online service from OpenAI and the Codex tool
- Input length
- Varies by product and mode
- Inputs
- Text and, depending on the service used, image inputs
Data checked . Release date unverified.
Indicative capability profile
Where the model is strong
The five levels are our clear summary of the results below. They are not a ranking that applies to every task.
Ideal use
When to choose it
- Everyday programming with AI assistance
- Analyses where costs need to be controlled
- Second stage after a low-cost classification model
Usage boundaries
When to choose another model
- The hardest tasks with no opportunity for correction
- Simple processing at scale where Flash-Lite is sufficient
Traceable supporting sources
Data and measurement sources
Distinguish between the manufacturer's documentation and the results of a specific test. Measurements also depend on the settings and the task set used.
LiveBench 2026-06-25
79.8 overall · 90.6 for problem solvingPerformance close to the most expensive variants at a lower cost per successful task.
Open original source ↗Agentic Coding
68.0 pointsIn this test, it performed better at autonomous coding than the Sol variant.
Open original source ↗Standard price after the price reduction
$2 input · $0.20 cached input · $12 outputAmounts are per million tokens in a short context; the price fell by 20%.
Open original source ↗Trends over time
Related events from AI Radar
OpenAI cuts prices for GPT-5.6 Luna and Terra models, says optimization of its own infrastructure by Sol made this possible
Anyone using OpenAI API for their own projects or tools with a low-cost model (Luna) will pay significantly less for the same work than before, as illustrated by Simon Willison, who switched from Gemini to Luna because of the price.
Companies running products on cheaper GPT-5.6 models can reduce inference costs by as much as tens of percent or consider switching from competing providers (Google, Anthropic), intensifying price pressure across the AI industry.
Chatham Financial uses the Codex tool and the GPT-5.6 model to reduce trade validation time
According to OpenAI, Chatham Financial reduced trade validation from 30 minutes to less than 4 minutes, which reduces the time needed for this step of the corporate process.
Company Basis: GPT-6 Astra model doubled the speed of tax workbook processing compared to the GPT-5.6 Sol model
For individuals working with large financial or tax spreadsheets, the case study suggests that a newer model may process similar tasks significantly faster, although this is a corporate testimonial, not a publicly available benchmark.
According to the company Basis, deploying the GPT-6 Astra model doubled the processing speed of a large 50-tab tax workbook compared to the GPT-5.6 Sol model, which increases the company's confidence in deploying the model for similar tasks in production.
Anthropic has released Claude Opus 5 – a cheaper model close to Fable 5 in performance
Claude Pro and Max users get Opus 5 as the default or most capable available model with no price change; the choice of "effort" level (low to max) directly affects the balance of speed, quality, and cost for individual queries.
Companies deploying Claude in products or development gain a model that, according to Anthropic, comes close to Fable 5 in knowledge work and agentic coding at half the token price, changing the cost calculation for running AI.
Next step
Compare the model using your actual task.
A benchmark narrows the selection. A short trial on your data determines the choice.