GPT-6 Sol in OffloopGive everyday work a clear finish line.
GPT Terra in Offloop currently runs GPT-6 Sol. Use it for bounded coding, document work and repeatable agent tasks.
A brief to try · Example
Handle a scoped code change
Fix this issue without changing adjacent behavior. Show the affected files and explain how the fix is verified.
Choose in your agent settings: GPT Terra
Where GPT-6 Sol fits
Our pick for well-specified tasks whose output you can inspect directly.
01
Handle a scoped code change
A bounded patch gives the model an explicit definition of done.
02
Run a repeatable business workflow
Published AutomationBench results support evaluating it for tool-driven work.
03
Edit against an acceptance checklist
Independent testing highlights the need to inspect the final output.
Our editorial assessment · GPT Terra is the Offloop selection name; this page describes GPT-6 Sol, not GPT-6.1 Terra. Check completed artifacts rather than relying on a completion message.
Read the capability source ↗Watch a real evaluation
A useful counterpoint to launch scores: the creator reports that website-editing completion claims did not match the work they inspected.
Superbash (BoxminingAI) · Independent creator · External setup, not Offloop
Published evaluations
Selected results for GPT-6 Sol. Source checked November 5, 2026; these are not Offloop measurements.
DeepSWE v1.1 · max
68.8%
OpenAI report; long-horizon engineering tasks.
AutomationBench 1.0.6 · xhigh
33.2%
End-to-end workflows across 47 tools; provider evaluation.
Agents’ Last Exam V1 · max
56.4%
Professional workflows; effort and harness affect results.
Different task sets, model snapshots and harnesses are not directly comparable. Scores are evidence for a trial, not a guarantee of your result.
Before you start
Give your agent a real task.
Choose GPT Terra, connect the tools you need, and define what a useful result looks like.