diff --git a/read.go b/read.go
index e8f98116..167c94b4 100644
--- a/read.go
+++ b/read.go
@@ -21,6 +21,7 @@ import (
 	"hash/crc64"
 	"log"
 	"os"
+	"slices"
 	"sort"
 
 	"github.com/rs/xid"
@@ -94,20 +95,15 @@ func (r *reader) Str() (string, error) {
 }
 
 func (r *reader) readTOC(toc *indexTOC) error {
-	sz, err := r.r.Size()
-	if err != nil {
-		return err
-	}
-	r.off = sz - 8
-
-	var tocSection simpleSection
-	if err := tocSection.read(r); err != nil {
-		return err
-	}
-
-	r.seek(tocSection.off)
+	return r.readTOCSections(toc, []string{})
+}
 
-	sectionCount, err := r.U32()
+// readTOCSections reads the table of contents of the index file.
+//
+// If the tags parameter is non-empty, it reads only those tagged sections for efficiency
+// and does not populate the other sections.
+func (r *reader) readTOCSections(toc *indexTOC, tags []string) error {
+	tocSection, sectionCount, err := r.readHeader()
 	if err != nil {
 		return err
 	}
@@ -126,11 +122,14 @@ func (r *reader) readTOC(toc *indexTOC) error {
 				return err
 			}
 
+			skipSection := len(tags) > 0 && !slices.Contains(tags, tag)
 			sec := secs[tag]
 			if sec == nil || sec.kind() != sectionKind(kind) {
 				// If we don't recognize the section, we may be reading a newer index than the current version. Use
 				// a "dummy section" struct to skip over it.
+				skipSection = true
 				log.Printf("encountered malformed index section (%s), skipping over it", tag)
+
 				switch sectionKind(kind) {
 				case sectionKindSimple:
 					sec = &simpleSection{}
@@ -143,8 +142,14 @@ func (r *reader) readTOC(toc *indexTOC) error {
 				}
 			}
 
-			if err := sec.read(r); err != nil {
-				return err
+			if skipSection {
+				if err := sec.skip(r); err != nil {
+					return err
+				}
+			} else {
+				if err := sec.read(r); err != nil {
+					return err
+				}
 			}
 		}
 	} else {
@@ -169,6 +174,27 @@ func (r *reader) readTOC(toc *indexTOC) error {
 	return nil
 }
 
+func (r *reader) readHeader() (simpleSection, uint32, error) {
+	sz, err := r.r.Size()
+	if err != nil {
+		return simpleSection{}, 0, err
+	}
+	r.off = sz - 8
+
+	var tocSection simpleSection
+	if err := tocSection.read(r); err != nil {
+		return simpleSection{}, 0, err
+	}
+
+	r.seek(tocSection.off)
+
+	sectionCount, err := r.U32()
+	if err != nil {
+		return simpleSection{}, 0, err
+	}
+	return tocSection, sectionCount, nil
+}
+
 func (r *indexData) readSectionBlob(sec simpleSection) ([]byte, error) {
 	return r.file.Read(sec.off, sec.sz)
 }
@@ -573,10 +599,10 @@ func NewSearcher(r IndexFile) (Searcher, error) {
 func ReadMetadata(inf IndexFile) ([]*Repository, *IndexMetadata, error) {
 	rd := &reader{r: inf}
 	var toc indexTOC
-	if err := rd.readTOC(&toc); err != nil {
+	err := rd.readTOCSections(&toc, []string{"metaData", "repoMetaData"})
+	if err != nil {
 		return nil, nil, err
 	}
-
 	return rd.parseMetadata(toc.metaData, toc.repoMetaData)
 }
 
diff --git a/read_test.go b/read_test.go
index 9e7acd13..c230ed9c 100644
--- a/read_test.go
+++ b/read_test.go
@@ -32,7 +32,6 @@ import (
 
 	"github.com/google/go-cmp/cmp"
 	"github.com/google/go-cmp/cmp/cmpopts"
-
 	"github.com/sourcegraph/zoekt/query"
 )
 
@@ -467,3 +466,32 @@ func TestEncodeRanks(t *testing.T) {
 		return true
 	}, nil)
 }
+
+func BenchmarkReadMetadata(b *testing.B) {
+	file, err := os.Open("testdata/benchmark/zoekt_v16.00000.zoekt")
+	if err != nil {
+		b.Fatalf("Failed to open test file: %v", err)
+	}
+	defer file.Close()
+
+	indexFile, err := NewIndexFile(file)
+	if err != nil {
+		b.Fatalf("could not open index: %v", err)
+	}
+
+	b.ReportAllocs()
+	b.ResetTimer()
+
+	for i := 0; i < b.N; i++ {
+		repos, metadata, err := ReadMetadata(indexFile)
+		if err != nil {
+			b.Fatalf("ReadMetadata failed: %v", err)
+		}
+		if len(repos) != 1 {
+			b.Fatalf("expected 1 repository")
+		}
+		if metadata == nil {
+			b.Fatalf("expected non-nil metadata")
+		}
+	}
+}
diff --git a/section.go b/section.go
index c686faaa..adb758e3 100644
--- a/section.go
+++ b/section.go
@@ -127,8 +127,12 @@ func (s *simpleSection) end(w *writer) {
 // section is a range of bytes in the index file.
 type section interface {
 	read(*reader) error
+	// skip advances over the data in the section without reading it.
+	// NOTE: the section will not contain valid data after this call, and it should not be used.
+	skip(*reader) error
 	write(*writer)
-	kind() sectionKind // simple or complex, used in serialization
+	// kind encodes whether the section is simple or compound, and is used in serialization
+	kind() sectionKind
 }
 
 type sectionKind int
@@ -156,10 +160,17 @@ func (s *simpleSection) read(r *reader) error {
 		return err
 	}
 	s.sz, err = r.U32()
+	return err
+}
+
+func (s *simpleSection) skip(r *reader) error {
+	var err error
+	_, err = r.U32()
 	if err != nil {
 		return err
 	}
-	return nil
+	_, err = r.U32()
+	return err
 }
 
 func (s *simpleSection) write(w *writer) {
@@ -215,6 +226,18 @@ func (s *compoundSection) read(r *reader) error {
 	return err
 }
 
+func (s *compoundSection) skip(r *reader) error {
+	if err := s.data.skip(r); err != nil {
+		return err
+	}
+	if err := s.index.read(r); err != nil {
+		return err
+	}
+
+	_, err := r.r.Read(s.index.off, s.index.sz)
+	return err
+}
+
 // relativeIndex returns the relative offsets of the items (first
 // element is 0), plus a final marking the end of the last item.
 func (s *compoundSection) relativeIndex() []uint32 {