GitLab reports that GPT-6 Astra completes typical runs up to 43.4% faster than GPT-5.6 Sol and uses 42.7% fewer tokens per run on the GitLab Duo Agent Platform.
The model is now available on Duo Agent Platform, with evidence from GitLab’s internal evaluation showing faster performance and lower token consumption across benchmark tasks.
PERFORMANCE SNAPSHOT
In internal testing, GPT-6 Astra finished a typical run 43.4% faster than GPT-5.6 Sol and used 42.7% fewer tokens per run, while completing every task in the benchmark. TechStaged has also covered GitLab Credits Bring Usage-Based Pricing to Duo Agent Platform.
- Typical run faster by 43.4%
- Token usage reduced by 42.7% per run
- All benchmark tasks completed without stalls
WHAT THIS MEANS FOR TEAMS
The performance gains translate to faster agentic work for common tasks such as dependency updates, build fixes, and small multi-file changes. Shorter runs can reduce wait times and help teams manage larger backlogs more efficiently.
- Faster dependency updates
- Quicker build fixes
- More efficient backlogs across teams
DEPLOYMENT AND GOVERNANCE
GitaLab notes that admins can set the model per feature in model selection, ensuring each workload runs on the chosen model.
Governance remains consistent, with every model going through the same context layer, policy checks, and audit trail.
AVAILABILITY AND NEXT STEPS
GPT-6 Astra is available now on GitLab Duo Agent Platform and runs on GitLab Credits.
- New to Duo Agent Platform? Start a free trial.
- GitLab Premium or Ultimate subscribers can turn on Duo Agent Platform and use included GitLab Credits.
CONTEXT AND CAVEATS
GitLab’s evaluation notes that Astra completed 100% of benchmark tasks, with 63.3% of tasks returning a change that passed the benchmark tests, compared to 76.7% for GPT-5.6 Sol. Misses are returned as patches to correct rather than as empty runs.
The slower tail of runs at the 95th percentile was faster with Astra (49.2% faster than Sol). Astra also achieved similar performance to reduce token usage, contributing to broader token-budget efficiency across backlogs.
RELATED COVERAGE
SOURCES
- GitLab: GPT-6 Astra on GitLab: Faster runs, fewer tokens used Published · Primary source
