diff --git a/conversion/__init__.py b/conversion/__init__.py
index 5aad203..feec11c 100644
--- a/conversion/__init__.py
+++ b/conversion/__init__.py
@@ -262,6 +262,7 @@ MMPROJ_MODEL_MAP: dict[str, str] = {
     "Gemma4UnifiedForConditionalGeneration": "gemma",
     "Glm4vForConditionalGeneration": "qwen3vl",
     "Glm4vMoeForConditionalGeneration": "qwen3vl",
+    "Glm5vForConditionalGeneration": "kimivl",
     "GlmOcrForConditionalGeneration": "qwen3vl",
     "GlmasrModel": "ultravox",
     "Granite4VisionForConditionalGeneration": "granite",
diff --git a/conversion/kimivl.py b/conversion/kimivl.py
index 63b8a07..d9aa212 100644
--- a/conversion/kimivl.py
+++ b/conversion/kimivl.py
@@ -152,3 +152,30 @@ class KimiK25Model(MmprojModel):
             name = name.replace(".proj.2.", ".proj.linear_2.")
 
         yield from super().modify_tensors(data_torch, name, bid)
+
+
+@ModelBase.register("Glm5vForConditionalGeneration")
+class Glm5vModel(KimiK25Model):
+    """GLM-5.2-Vision: MoonViT3d encoder reused verbatim, projector maps to 6144"""
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+
+        # in_patch_limit sits under media_proc_cfg here, not at the top level
+        media_proc_cfg = self.preprocessor_config.get("media_proc_cfg")
+        if media_proc_cfg is not None:
+            self.preprocessor_config = {**media_proc_cfg, **self.preprocessor_config}
+
+    def set_gguf_parameters(self):
+        super().set_gguf_parameters()
+
+        # encoder graph is shared with kimik25; the type differs only so that
+        # mtmd wraps the embeddings in GLM's image delimiters
+        self.gguf_writer.add_clip_projector_type(gguf.VisionProjectorType.GLM5V)
+
+    def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
+        # projector is flat here (mm_projector.linear_N) rather than nested under proj
+        if name.startswith("mm_projector.linear_"):
+            name = name.replace("mm_projector.linear_", "mm_projector.proj.linear_")
+
+        yield from super().modify_tensors(data_torch, name, bid)
diff --git a/gguf-py/gguf/constants.py b/gguf-py/gguf/constants.py
index 1bda945..d9b58ed 100644
--- a/gguf-py/gguf/constants.py
+++ b/gguf-py/gguf/constants.py
@@ -4539,6 +4539,7 @@ class VisionProjectorType:
     KIMIVL = "kimivl"
     PADDLEOCR = "paddleocr"
     KIMIK25 = "kimik25"
+    GLM5V = "glm5v"
     LIGHTONOCR = "lightonocr"
     COGVLM = "cogvlm"
     JANUS_PRO = "janus_pro"
diff --git a/tools/mtmd/clip-impl.h b/tools/mtmd/clip-impl.h
index 5b41368..8897cd8 100644
--- a/tools/mtmd/clip-impl.h
+++ b/tools/mtmd/clip-impl.h
@@ -364,6 +364,7 @@ enum projector_type {
     PROJECTOR_TYPE_YOUTUVL,
     PROJECTOR_TYPE_YASA2,
     PROJECTOR_TYPE_KIMIK25,
+    PROJECTOR_TYPE_GLM5V,
     PROJECTOR_TYPE_NEMOTRON_V2_VL,
     PROJECTOR_TYPE_HUNYUANVL,
     PROJECTOR_TYPE_EXAONE4_5,
@@ -418,6 +419,7 @@ static std::map<projector_type, std::string> PROJECTOR_TYPE_NAMES = {
     { PROJECTOR_TYPE_YOUTUVL,           "youtuvl"},
     { PROJECTOR_TYPE_YASA2,             "yasa2"},
     { PROJECTOR_TYPE_KIMIK25,           "kimik25"},
+    { PROJECTOR_TYPE_GLM5V,             "glm5v"},
     { PROJECTOR_TYPE_NEMOTRON_V2_VL,    "nemotron_v2_vl"},
     { PROJECTOR_TYPE_EXAONE4_5,         "exaone4_5"},
     { PROJECTOR_TYPE_HUNYUANVL,         "hunyuanvl"},
diff --git a/tools/mtmd/clip.cpp b/tools/mtmd/clip.cpp
index d2226b3..ff3a73f 100644
--- a/tools/mtmd/clip.cpp
+++ b/tools/mtmd/clip.cpp
@@ -957,6 +957,7 @@ static std::unique_ptr<clip_graph> clip_get_graph_builder(clip_ctx * ctx, const
                 builder = std::make_unique<clip_graph_paddleocr>(ctx, img);
             } break;
         case PROJECTOR_TYPE_KIMIK25:
+        case PROJECTOR_TYPE_GLM5V:
             {
                 builder = std::make_unique<clip_graph_kimik25>(ctx, img);
             } break;
@@ -1399,6 +1400,7 @@ struct clip_model_loader {
                         hparams.set_warmup_n_tokens(256); // avoid OOM on warmup
                     } break;
                 case PROJECTOR_TYPE_KIMIK25:
+                case PROJECTOR_TYPE_GLM5V:
                     {
                         hparams.image_resize_algo = RESIZE_ALGO_BICUBIC;
                         hparams.rope_theta = 10000.0f;
@@ -2307,6 +2309,7 @@ struct clip_model_loader {
             case PROJECTOR_TYPE_KIMIVL:
             case PROJECTOR_TYPE_PADDLEOCR:
             case PROJECTOR_TYPE_KIMIK25:
+            case PROJECTOR_TYPE_GLM5V:
                 {
                     model.mm_input_norm_w = get_tensor(TN_MM_INP_NORM);
                     model.mm_input_norm_b = get_tensor(TN_MM_INP_NORM_B);
@@ -3381,6 +3384,7 @@ int clip_n_output_tokens(const clip_ctx * ctx, const clip_image_f32 * img) {
         case PROJECTOR_TYPE_LFM2:
         case PROJECTOR_TYPE_KIMIVL:
         case PROJECTOR_TYPE_KIMIK25:
+        case PROJECTOR_TYPE_GLM5V:
             {
                 // dynamic size
                 int out_patch_size = params.patch_size * ctx->model.hparams.n_merge;
@@ -4040,6 +4044,7 @@ bool clip_image_batch_encode(clip_ctx * ctx, int n_threads, const clip_image_f32
         case PROJECTOR_TYPE_PIXTRAL:
         case PROJECTOR_TYPE_KIMIVL:
         case PROJECTOR_TYPE_KIMIK25:
+        case PROJECTOR_TYPE_GLM5V:
         case PROJECTOR_TYPE_LIGHTONOCR:
             {
                 // set the 2D positions
@@ -4588,6 +4593,7 @@ int clip_n_mmproj_embd(const struct clip_ctx * ctx) {
         case PROJECTOR_TYPE_KIMIVL:
         case PROJECTOR_TYPE_PADDLEOCR:
         case PROJECTOR_TYPE_KIMIK25:
+        case PROJECTOR_TYPE_GLM5V:
         case PROJECTOR_TYPE_YASA2:
             return ctx->model.mm_2_w->ne[1];
         case PROJECTOR_TYPE_HUNYUANVL:
diff --git a/tools/mtmd/mtmd.cpp b/tools/mtmd/mtmd.cpp
index 724538b..bd44fd4 100644
--- a/tools/mtmd/mtmd.cpp
+++ b/tools/mtmd/mtmd.cpp
@@ -560,6 +560,13 @@ struct mtmd_context {
                     img_end = "<|media_end|>";
                     image_preproc = std::make_unique<mtmd_image_preprocessor_dyn_size>(ctx_v);
                 } break;
+            case PROJECTOR_TYPE_GLM5V:
+                {
+                    // same encoder as kimik25, but GLM's vocab has no <|media_*|>
+                    img_beg = "<|begin_of_image|>";
+                    img_end = "<|end_of_image|>";
+                    image_preproc = std::make_unique<mtmd_image_preprocessor_dyn_size>(ctx_v);
+                } break;
             case PROJECTOR_TYPE_LIGHTONOCR:
                 {
                     // <|im_start|> ... (image embeddings) ... <|im_end|>