DSPy GEPA: reflective prompt optimization
A cookbook published on October 16, 2025 reports an 11% accuracy increase on NuminaMath-1.5, at a total cost below US$0.50.
Source: Otimização de prompts com DSPy GEPA (huggingface.co). Text prepared with AI from this source.
What happened and what to do
Published on October 16, 2025, the cookbook presents reflective prompt optimization with DSPy GEPA and an example using separate models for inference and reflection. The publication reports an 11% accuracy increase on NuminaMath-1.5 for less than US$0.50 total. To check the scope and findings, consult the original cookbook by searching for the title “Otimização de prompts com DSPy GEPA” and review its stated conditions and metrics.
A company can apply this evaluation approach to its own tasks: define metrics and test sets, compare prompt versions, and track results, costs, and regressions. Any implementation should validate performance in its actual context before adopting changes; AI need not be introduced where conventional processes better meet the need.
How the consultancy can help
Wendelmaques can assess the prompt evaluation process and its data, define an implementation scope for testing, metrics, and monitoring, and propose operations suited to the company’s infrastructure.
Next step
Send a short description of the task, available data, and desired outcome to receive a scoped proposal.
Consulting for your project
Infrastructure review, deployment and ongoing operations, with scope and pricing defined in the proposal.
Quoted per project
Request a proposal