diff --git a/cmd/gguf-parser/main.go b/cmd/gguf-parser/main.go
index 598452f..93f4b67 100644
--- a/cmd/gguf-parser/main.go
+++ b/cmd/gguf-parser/main.go
@@ -1315,6 +1315,7 @@ func mainAction(c *cli.Context) error {
 				"Flash Attention",
 				"MMap Load",
 				"Embedding Only",
+				"Reranking",
 				"Distributable",
 				"Offload Layers",
 				"Full Offloaded",
@@ -1326,6 +1327,7 @@ func mainAction(c *cli.Context) error {
 				"Flash Attention",
 				"MMap Load",
 				"Embedding Only",
+				"Reranking",
 				"Distributable",
 				"Offload Layers",
 				"Full Offloaded",
@@ -1385,6 +1387,7 @@ func mainAction(c *cli.Context) error {
 					sprintf(tenary(flashAttention, tenary(es.FlashAttention, "Enabled", "Unsupported"), "Disabled")),
 					sprintf(tenary(mmap, tenary(!es.NoMMap, "Enabled", "Unsupported"), "Disabled")),
 					sprintf(tenary(es.EmbeddingOnly, "Yes", "No")),
+					sprintf(tenary(es.Reranking, "Supported", "Unsupported")),
 					sprintf(tenary(es.Distributable, "Supported", "Unsupported")),
 					sprintf(tenary(es.Items[i].FullOffloaded, sprintf("%d (%d + 1)",
 						es.Items[i].OffloadLayers, es.Items[i].OffloadLayers-1), es.Items[i].OffloadLayers)),
diff --git a/file_estimate.go b/file_estimate.go
index 93b25fa..65fef70 100644
--- a/file_estimate.go
+++ b/file_estimate.go
@@ -34,6 +34,9 @@ type (
 		// EmbeddingOnly is the flag to indicate whether the model is used for embedding only,
 		// true for embedding only.
 		EmbeddingOnly bool `json:"embeddingOnly"`
+		// Reranking is the flag to indicate whether the model is used for reranking,
+		// true for reranking.
+		Reranking bool `json:"reranking"`
 		// Distributable is the flag to indicate whether the model is distributable,
 		// true for distributable.
 		Distributable bool `json:"distributable"`
@@ -215,6 +218,10 @@ func (gf *GGUFFile) EstimateLLaMACppRun(opts ...LLaMACppRunEstimateOption) (e LL
 	if a.Type == "model" && !a.AttentionCausal {
 		e.EmbeddingOnly = true
 		o.PhysicalBatchSize = o.LogicalBatchSize
+		// Reranking.
+		if _, found := gf.TensorInfos.Index([]string{"cls.bias", "cls.weight"}); found > 0 {
+			e.Reranking = true
+		}
 	}
 
 	// Distributable,
@@ -357,6 +364,9 @@ func (gf *GGUFFile) EstimateLLaMACppRun(opts ...LLaMACppRunEstimateOption) (e LL
 	ls := gf.Layers()
 	ioLs, tfLs, _ := ls.Cut([]string{
 		"token_embd.weight",
+		"token_embd_norm.weight",
+		"token_embd_norm.bias",
+		"token_types.weight",
 		"output.weight",
 		"output.bias",
 		"output_norm.weight",
@@ -364,6 +374,9 @@ func (gf *GGUFFile) EstimateLLaMACppRun(opts ...LLaMACppRunEstimateOption) (e LL
 	})
 	ipLs, opLs, _ := ioLs.Cut([]string{
 		"token_embd.weight",
+		"token_embd_norm.weight",
+		"token_embd_norm.bias",
+		"token_types.weight",
 	})
 
 	// Weight.
@@ -685,6 +698,9 @@ type (
 		// EmbeddingOnly is the flag to indicate whether the model is used for embedding only,
 		// true for embedding only.
 		EmbeddingOnly bool `json:"embeddingOnly"`
+		// Reranking is the flag to indicate whether the model is used for reranking,
+		// true for reranking.
+		Reranking bool `json:"reranking"`
 		// Distributable is the flag to indicate whether the model is distributable,
 		// true for distributable.
 		Distributable bool `json:"distributable"`
@@ -848,6 +864,7 @@ func (e LLaMACppRunEstimate) Summarize(mmap bool, nonUMARamFootprint, nonUMAVram
 	es.FlashAttention = e.FlashAttention
 	es.NoMMap = e.NoMMap
 	es.EmbeddingOnly = e.EmbeddingOnly
+	es.Reranking = e.Reranking
 	es.LogicalBatchSize = e.LogicalBatchSize
 	es.PhysicalBatchSize = e.PhysicalBatchSize
 	es.Distributable = e.Distributable