gpt-5.6-terra
Meant to excel at
No sourced claims yet.
Watch out for
No sourced claims yet.
Benchmarks
Agentic coding percentile 0.67
| Benchmark | Score | Percentile | Source | Date |
|---|---|---|---|---|
| artificial analysis coding agent index | 77 | 0.80 | Independent evaluator artificialanalysis.ai | 2026-07-09 |
| terminal-bench 2.1 (aggregator, anotherwrapper.com) | 87.4 | 0.54 | Independent evaluator anotherwrapper.com | 2026-10-19 |
Percentile is the share of tracked models this model beats on that benchmark, averaged across the group. Only benchmarks at least four tracked models share are scored.
Practitioner reports
- +
Says Terra Ultra works well in Codex when tasks are explained in detail, claiming roughly twice the quality and speed of GPT-5.5.
Task: feature | Compared with gpt-5.5 | 2026-07-14 older than 90 days
Practitioner report community.openai.com - -
A Plus user on Terra at medium says the whole weekly Codex allowance was gone in about six hours of active development, instructions were followed early then gradually forgotten, and the model made changes outside the requested scope.
Task: agentic-long-running | Compared with gpt-5.6-sol | 2026-07-13 older than 90 days
Practitioner report community.openai.com - -
Reports GPT-5.4 at medium beat 5.6 Terra at story writing, with Terra having trouble tracking objects and character knowledge.
Task: writing-content | Compared with gpt-5.4 | 2026-07-13 older than 90 days
Practitioner report community.openai.com - +
In a creator review roundup, Terra was preferred over Sol for writing direct, concise PRDs.
Task: writing-content | Compared with gpt-5.6-sol | 2026-07-12 older than 90 days
Practitioner report ai.joaoqueiros.com - -
In CodeRabbit's tests Terra solved 40.7% of coding tasks versus 63.7% for Sol while using about 2.6x the output tokens per task, and was weaker at review; CodeRabbit sees it as fit for bounded repeatable work and first-pass triage, not as a cost saver by default.
Task: feature | Compared with gpt-5.6-sol | 2026-07-09 older than 90 days
Practitioner report coderabbit.ai - -
A Codex Reddit thread (relayed secondhand by an aggregator) says Terra burned nearly as much usage as Sol without being better, and that Sol at a lower setting is the better value; Terra described as poorly positioned.
Task: feature | Compared with gpt-5.6-sol undated
Practitioner report go.tabbit.ai