diff --git a/src/models/local/llamacpp.test.ts b/src/models/local/llamacpp.test.ts
index 9812c90..8e0a91d 100644
--- a/src/models/local/llamacpp.test.ts
+++ b/src/models/local/llamacpp.test.ts
@@ -24,6 +24,7 @@ describe('LlamaCppClient', () => {
 
     const client = new LlamaCppClient({
       endpoint: 'http://localhost:8080',
+      model: 'test-model',
     });
 
     const response = await client.chat({
@@ -64,6 +65,7 @@ describe('LlamaCppClient', () => {
 
     const client = new LlamaCppClient({
       endpoint: 'http://localhost:8080',
+      model: 'test-model',
     });
 
     const events: ChatStreamEvent[] = [];
@@ -87,6 +89,7 @@ describe('LlamaCppClient', () => {
 
     const client = new LlamaCppClient({
       endpoint: 'http://localhost:8080',
+      model: 'test-model',
     });
 
     await expect(client.chat({
diff --git a/src/models/local/llamacpp.ts b/src/models/local/llamacpp.ts
index f428b38..6665d1d 100644
--- a/src/models/local/llamacpp.ts
+++ b/src/models/local/llamacpp.ts
@@ -2,6 +2,7 @@ import type { ChatRequest, ChatResponse, ChatStreamEvent, ModelClient } from '..
 
 export interface LlamaCppClientConfig {
   endpoint: string;
+  model: string;
   authToken?: string;
 }
 
@@ -22,10 +23,12 @@ interface LlamaCppStreamChunk {
 
 export class LlamaCppClient implements ModelClient {
   private endpoint: string;
+  private model: string;
   private authToken?: string;
 
   constructor(config: LlamaCppClientConfig) {
     this.endpoint = config.endpoint.replace(/\/$/, '');
+    this.model = config.model;
     this.authToken = config.authToken;
   }
 
@@ -54,6 +57,7 @@ export class LlamaCppClient implements ModelClient {
         method: 'POST',
         headers,
         body: JSON.stringify({
+          model: this.model,
           messages,
           max_tokens: request.maxTokens ?? 2048,
         }),
@@ -106,6 +110,7 @@ export class LlamaCppClient implements ModelClient {
         method: 'POST',
         headers,
         body: JSON.stringify({
+          model: this.model,
           messages,
           max_tokens: request.maxTokens ?? 2048,
           stream: true,