diff --git a/crates/gc-extract/tests/disc.rs b/crates/gc-extract/tests/disc.rs
index c227c1d..a67606b 100644
--- a/crates/gc-extract/tests/disc.rs
+++ b/crates/gc-extract/tests/disc.rs
@@ -4045,6 +4045,7 @@ fn sound_bank_sblk_head_is_constant_apart_from_one_word() {
     let mut values: Vec<u32> = Vec::new();
     let mut bodies: Vec<usize> = Vec::new();
     let mut levels = 0usize;
+    let mut records_seen = 0usize;
     for lv in 0..27usize {
         let Ok(bytes) = std::fs::read(data.join(format!("level{lv}/sound_bank.bin"))) else {
             continue;
@@ -5731,6 +5732,7 @@ fn moby_class_ids_come_from_lvl_vtbl_not_from_the_class_count() {
     let mut classes: BTreeSet<u32> = BTreeSet::new();
     let mut records = 0usize;
     let mut levels = 0usize;
+    let mut records_seen = 0usize;
     for lv in 0..27usize {
         let Ok(ov) = std::fs::read(data.join(format!("level{lv}/overlay.bin"))) else {
             continue;
@@ -5780,3 +5782,114 @@ fn moby_class_ids_come_from_lvl_vtbl_not_from_the_class_count() {
         "the small ids that open every vtbl are missing"
     );
 }
+
+/// `lvl.vtbl`'s `oClass` is a **global** moby class identifier, not an index into
+/// the level's class table.
+///
+/// The two are easy to conflate and are not the same axis. Across the 27 levels
+/// the vtbl carries 1,163 distinct identifiers reaching **5,050**, while
+/// `moby_classes.count` runs 161 to 259 per level. So `oClass` cannot index the
+/// class table -- almost every identifier would fall off the end.
+///
+/// The identifiers are *shared* rather than per-level: the set is cumulative
+/// across the 27 levels and no level re-uses the numbering, so 1,163 distinct
+/// identifiers come out of 5,878 records. Level 0 contributes 234 and levels 1-8
+/// contribute 72, 69, 77, 54, 41, 68 and 28 -- roughly each level's own share,
+/// not a single level dominating.
+///
+/// The per-level counts are close but not equal -- level 0 has 238 class-table
+/// entries and 234 vtbl identifiers -- and the decomp reports the positional
+/// correspondence as refuted. Nothing here depends on that: the useful fact is
+/// that the identifier space is global and known, so a class can be recognised
+/// across levels by name.
+#[test]
+fn vtbl_class_ids_are_global_and_do_not_index_the_class_table() {
+    require_disc!();
+
+    let data = std::env::var_os("GC_EXTRACTED")
+        .map(PathBuf::from)
+        .unwrap_or_else(|| {
+            let home = std::env::var("HOME").unwrap_or_else(|_| ".".to_string());
+            PathBuf::from(home).join("Documents/gc-extracted")
+        });
+    if !data.exists() {
+        eprintln!("skipped: unpack the disc first");
+        return;
+    }
+
+    use gc_formats::level_core::LevelCoreHeader;
+    use gc_formats::overlay_reloc::parse_overlay_sections;
+    use std::collections::BTreeSet;
+
+    let mut ids: BTreeSet<u32> = BTreeSet::new();
+    let mut first_level_new = 0usize;
+    let mut per_level_new = Vec::new();
+    let mut max_count = 0u32;
+    let mut min_count = u32::MAX;
+    let mut levels = 0usize;
+    let mut records_seen = 0usize;
+    for lv in 0..27usize {
+        let (Ok(idx), Ok(ov)) = (
+            std::fs::read(data.join(format!("level{lv}/core_index.bin"))),
+            std::fs::read(data.join(format!("level{lv}/overlay.bin"))),
+        ) else {
+            continue;
+        };
+        let count = LevelCoreHeader::parse(&idx)
+            .expect("hdr")
+            .moby_classes
+            .count as u32;
+        max_count = max_count.max(count);
+        min_count = min_count.min(count);
+        let secs = parse_overlay_sections(&ov).expect("sections");
+        let before = ids.len();
+        for r in 0..secs[3].data.len() / 12 {
+            let c = u32::from_le_bytes(secs[3].data[r * 12..r * 12 + 4].try_into().unwrap());
+            if c != u32::MAX {
+                ids.insert(c);
+                records_seen += 1;
+            }
+        }
+        let added = ids.len() - before;
+        if lv == 0 {
+            first_level_new = added;
+        }
+        per_level_new.push(added);
+        levels += 1;
+    }
+    assert_eq!(levels, 27, "only {levels} levels parsed");
+
+    let largest = *ids.iter().next_back().unwrap();
+    let out_of_range = ids.iter().filter(|&&c| c >= max_count).count();
+    eprintln!(
+        "vtbl oClass: {} ids, largest {largest}; moby_classes.count runs \
+         {min_count}..{max_count}; {out_of_range}/{} ids are at or above the \
+         largest per-level table; level 0 alone contributes {first_level_new}",
+        ids.len(),
+        ids.len()
+    );
+
+    assert_eq!(ids.len(), 1163, "distinct vtbl class ids moved");
+    assert_eq!(largest, 5050, "the largest class id moved");
+    // the load-bearing claim: the identifier space does not fit in a level's table
+    assert!(
+        largest > max_count * 10,
+        "the id space now fits inside a class table; the two axes may have merged"
+    );
+    assert!(
+        out_of_range * 10 >= ids.len() * 9,
+        "most ids should be out of range for any per-level table"
+    );
+    // and it is shared, not per-level: 1,163 identifiers come out of 5,878
+    // records, so levels re-use ids rather than renumbering. No single level
+    // dominates -- level 0's 234 is about a fifth of the total.
+    assert!(
+        ids.len() * 3 < records_seen,
+        "the id count is now close to the record count; the numbering may be per-level"
+    );
+    assert!(
+        first_level_new * 4 <= ids.len(),
+        "level 0 supplied {first_level_new} of {} ids -- that is dominance, not a share",
+        ids.len()
+    );
+}
diff --git a/docs/formats/core-data-regions.md b/docs/formats/core-data-regions.md
index e4633e1..457b196 100644
--- a/docs/formats/core-data-regions.md
+++ b/docs/formats/core-data-regions.md
@@ -3120,3 +3120,32 @@ with nothing above them is what a broad property looks like, not a field. The
 positive finding stands and is now usable: **the class-id space is known, and it
 is not the class count.** Asserted in
 `moby_class_ids_come_from_lvl_vtbl_not_from_the_class_count`.
+
+
+## `oClass` is a global identifier, and not an index into the class table
+
+Two axes that are easy to conflate, and are not the same thing:
+
+    lvl.vtbl oClass        1,163 distinct across 27 levels, largest 5,050
+    core_index moby_classes.count   136 .. 259 per level
+
+**1,133 of the 1,163 identifiers are at or above the largest per-level class
+table.** So `oClass` cannot index that table -- nearly every identifier would fall
+off the end. They are different axes: the vtbl names a *type* of moby, the class
+table lists the *models* a level instantiates.
+
+The identifier space is shared rather than renumbered per level: 1,163 distinct
+values come out of 5,878 records, and level 0 contributes 234 while levels 1-8
+contribute 72, 69, 77, 54, 41, 68 and 28 -- each level roughly its own share, not
+one level dominating. An earlier note here said "most of the game's classes appear
+in the first level"; the data says 20%, and the assertion is written to fail if
+that ever changes.
+
+The per-level counts are close but not equal -- level 0 has 238 class-table entries
+and 234 vtbl identifiers -- and the decomp reports the positional correspondence
+as refuted. Nothing here depends on that.
+
+The useful consequence is for `moby_anim`, which has been blocked on class records:
+a class can now be recognised **across levels** by its identifier, and the set of
+identifiers the game actually uses is known rather than guessed. Asserted in
+`vtbl_class_ids_are_global_and_do_not_index_the_class_table`.
diff --git a/docs/roadmap.md b/docs/roadmap.md
index 38df58a..197839d 100644
--- a/docs/roadmap.md
+++ b/docs/roadmap.md
@@ -407,6 +407,13 @@ Phase 1 hands over the extents; this phase works out what each one contains.
       the chunk class-ID artifact now retracted. Re-scored against the real set,
       no chunk slot separates (best 0.375 against 0.018 chance, a dozen slots in
       the same band). Asserted.
+- [x] `lvl.vtbl`'s **`oClass` is a global identifier**, not an index into the level's
+      class table: 1,163 distinct ids across the game, largest 5,050, against
+      `moby_classes.count` of 136..259. **1,133 of 1,163 are out of range** for any
+      per-level table. The two are different axes -- a type id, and the models a
+      level instantiates -- and the space is shared rather than renumbered
+      (1,163 distinct out of 5,878 records; level 0 supplies 20%, not a
+      majority). A class can now be recognised across levels by id. Asserted.
 - [ ] `moby_anim` — **still open, and the obvious reading is ruled out.** An
       instance table would have to scale with the level's moby class count, but
       **per-chunk-type record counts are level-independent** -- the same types