Mubit and Neon release agent template that cuts frontier-model tokens
Start here
Mubit and Neon published a template for a data-analyst agent that routes every model call through Mubit to Neon AI Gateway, aiming to cut frontier-model token spend. In Mubit's benchmark on 20 CL-bench questions run three times each, the setup reduced gpt-5.5 input tokens by 73% and answered 40% faster, with accuracy at 65% versus 60% for gpt-5.5 alone. The mechanism: after a small tuned model passes an exam built from the task's own traffic, Mubit hands it routine steps such as listing tables and writing up answers, while gpt-5.5 still writes every SQL query. The agent runs on Neon Functions and Lakebase Postgres. Using the template requires a Neon project with AI Gateway prepaid credits in a region supporting Neon Functions and AI Gateway, Node 24+ and the signed-in Neon CLI, and a Mubit project whose only provider key is a Neon AI Gateway key. The 73% and 40% figures come from Mubit's own benchmark, not independent testing.
Generated from reports · updated 13 h ago
Report timeline
Follow the reports to see the event from different sides.
- Hacker News · AINeon and Mubit release a template for an agent that cuts its own frontier-model tokens by 73%
Neon and Mubit published a template for a data analyst agent on Neon Functions and Lakebase Postgres that sends every model call through Mubit to Neon AI Gateway. In Mubit's benchmark on 20 CL-bench questions run three times each, the setup cut gpt-5.5 input tokens by 73% and answered 40% faster, with accuracy at 65% versus 60% for gpt-5.5 alone. Mubit moves routine steps such as listing tables and writing up answers to a small tuned model after it passes an exam built from the task's own traffic, while gpt-5.5 still writes every SQL query.
Heat for this event
Not enough continuous observations yet to draw a trend.