A January 13, 2026 publication reports the launch of FrogMini, a 14B coding agent, and gives a score of 45.3% on SWE-Bench Verified.
On January 13, 2026, a publication reported that Microsoft launched FrogMini, a 14B coding agent, on a public model platform. According to the report, the agent scored 45.3% on SWE-Bench Verified. The publication also describes a training approach in which agents can unintentionally break tests while adding features.
The reported score and approach may interest readers following coding-agent evaluation and training, but they do not, by themselves, establish performance in other settings. Consult the original publication and verify the benchmark methodology, setup, and results before drawing conclusions. If using AI to study or apply this material, avoid submitting proprietary code or personal data without authorization, and follow your organization’s policies.