GPT-6 Astra on GitLab: Faster runs, fewer tokens used

This post announces the integration of GPT-6 Astra into the GitLab Duo Agent Platform. It details internal benchmark results showing GPT-6 Astra is 43.4% faster and uses 42.7% fewer tokens per run than GPT-5.6 Sol. It highlights improvements in tail latency (49.2% faster at the 95th percentile) and a 100% task completion rate with a 63.3% resolution rate (compared to 76.7% for Sol). The post emphasizes Astra's suitability for tasks like dependency updates and build fixes where turnaround time and token usage are critical, while Sol might be preferred for changes requiring absolute correctness. It also reiterates that the platform provides consistent governance, context layering, policy checks, and audit trails for all integrated models.