Understand this story
On November 12, 2026, Baseten released Baseten Router, a model router for coding tasks. According to the Baseten engineering blog, the router selects between Opus 5.5 and GLM 5.3 Flash based on the coding task; on the DeepSWE 1.1, SWE Atlas QnA, and Terminal-Bench 2.1 test sets, its pass rate is only 0.8 percentage points behind Opus 5.5, while cost is reduced by 72%. It offers three threshold tiers—Cost, Balanced, and Intelligence—which can be switched at request time without retraining, and it is now in public preview as part of Baseten Grid. Deployed on a single B200 GPU, p50 and p95 latency at 4 concurrent requests are 273 ms and 429 ms, respectively.
AI-generated from public reports
Latest development · Report
Baseten Releases Baseten Router, a Model Router for Coding TasksDevelopments
1 confirmed developments
Baseten releases Baseten Router, a model router for coding tasks
Source:English text is being prepared.
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 12
Baseten Engineering BlogFirst-party
Baseten Releases Baseten Router, a Model Router for Coding Tasks
Baseten has released Baseten Router, which selects between Opus 5.5 and GLM 5.3 Flash based on the coding task. On the DeepSWE 1.1, SWE Atlas QnA, and Terminal-Bench 2.1 test sets, its pass rate is only 0.8 percentage points below Opus 5.5, while cost is reduced by 72%. The router offers three threshold tiers—Cost, Balanced, and Intelligence—which can be switched at request time without retraining, and it is now in public preview as part of Baseten Grid. Deployed on a single B200 GPU, p50 and p95 latency at 4 concurrent requests are 273 ms and 429 ms, respectively.