The "semi" seems to imply that free will is incompatible with determinism, otherwise, why distinguish it from compatibilism? But John Martin Fischer, who originated the term, says it has nothing to do with freedom.
Randolph Clarke suggested the term "narrow compatibilism" to distinguish it from the original "broad compatibilism" that makes determinism compatible with both free will and moral responsibility. For Clarke, narrow compatibilism is distinguished from broad by asserting we have moral responsibility, even if determinism is true and libertarian free will does not exist. Note that Fischer, Alfred Mele, and many others for whom moral responsibility is most important, are generally agnostic about the "truth" of either libertarian free will or determinism. They want moral responsibility even in the absence of "alternative possibilities" and the idea that an agent "could have done otherwise." Inspired by Peter Strawson, these thinkers affirm moral responsibility and accountability for their "autonomous agents" whatever the outcome of the problem of free will and determinism.
Let's look at the taxonomy of deterministic positions and see where semicompatibilism fits.
Recently, incompatibilists have staked out nuanced versions of the familiar positions with new jargon like hard incompatibilism, semicompatibilism, and illusionism.
Broad compatibilists think both free will and moral responsibility are compatible with determinism. Narrow compatibilists think free will is not compatible, but moral responsibility is compatible with determinism. Semicompatibilists are narrow compatibilists who are agnostic about free will and determinism. They think moral responsibility is compatible with determinism or indeterminism, if either is true.
Hard incompatibilists think both free will and moral responsibility are not compatible with determinism. Illusionists are incompatibilists who say free will is an illusion.
Soft incompatibilists think both free will and moral responsibility are incompatible with strict determinism, but both are compatible with an adequate determinism.