The timing of diamond crystallization is generally inferred from radiometric dating of individual or multiple mineral inclusions in diamond, where ages are derived from isochrons or isotopic evolution models. Resulting ages often significantly predate the emplacement ages of host kimberlites, but age information from both approaches can be ambiguous due to long-lived mantle heterogeneities where mixing can mimic isochrons or in-situ aging. Direct dating of accessory mineral inclusions in diamonds, by contrast, has rarely been attempted because of the scarcity of such inclusions, requiring careful high grading of large amounts of diamond concentrates. Here, we report secondary ionization mass spectrometry (SIMS) U–Pb ages from a suite of rare rutile inclusions which were extracted from diamond in an eclogitic paragenesis from the Mir kimberlite pipe, Siberia. One population of rutile inclusions shows diamond crystal-shapes and was petrographically identified as completely enclosed in diamond prior to host burning for inclusion extraction. These rutile inclusions were screened for U abundances (0.24–23 μg/g U) and those with the highest U contents were dated, yielding a uniform U–Pb concordia age of 375.5 ± 7.0 Ma (error 95% confidence; mean square of weighted deviates MSWD = 0.74; number of spots n = 19). This age is nearly coeval with, but significantly (at the 95% confidence level) older than combined U–Pb rutile ages for rutile intergrown with polycrystalline fibrous diamond (362.9 ± 9.5 Ma; MSWD = 0.96; n = 7) and rutile from the eclogitic xenolith matrix (369 ± 16 Ma; MSWD = 0.30; n = 5). Rutile ages are consistent with published ages for kimberlite emplacement (ca. 360 Ma), but much younger than Re–Os sulfide isochron ages between ca. 913 Ma and 2.1 Ga for groups of individual inclusions in diamond from the same location. When assuming the oldest Re–Os sulfide age as the timing of rutile entrapment, diffusive Pb-loss from rutile inclusions over ca. 2 Ga would require diffusivities of Pb in diamond of >10−22 m2/s, as well as Pb partitioning coefficients between diamond and rutile at near or greater unity. Alternatively, if Pb is immobile in diamond, ancient rutile ages could have been reset due to processes other than volume diffusion through diamond (e.g., due to fluid migration along cracks). However, in this scenario it remains difficult to explain why all rutile inclusions dated here would have been located along cracks, whereas sulfide inclusions investigated in other studies were not. Another explanation is that rutile inclusions were entrapped at the time of diamond formation, and that this event is recorded by U–Pb rutile ages, whereas sulfides predate diamond crystallization and their Re–Os systematics were incompletely reset. Trace element heterogeneity in rutile suggests that they originated in chemically different environments, and thus rutile likely predates diamond formation. However, due to rapid diffusion of Pb in rutile, radiogenic Pb accumulated only after entrapment in diamond (for inclusions) or eruptive quenching (for intergrown or matrix rutile). Overall, rutile geochronology indicates that at least some diamond from the Mir pipe formed briefly (within the ca. 20 Ma age difference between U–Pb rutile ages for inclusions and intergrown/matrix rutile) before kimberlite eruption, supporting models that link diamond formation with carbon-rich precursors of kimberlite magmas.