diff --git a/crates/forge_domain/src/provider.rs b/crates/forge_domain/src/provider.rs
index 3a978baf59..c77ce1b28f 100644
--- a/crates/forge_domain/src/provider.rs
+++ b/crates/forge_domain/src/provider.rs
@@ -70,6 +70,7 @@ impl ProviderId {
     pub const MINIMAX: ProviderId = ProviderId(Cow::Borrowed("minimax"));
     pub const CODEX: ProviderId = ProviderId(Cow::Borrowed("codex"));
     pub const OPENCODE_ZEN: ProviderId = ProviderId(Cow::Borrowed("opencode_zen"));
+    pub const NOVITA: ProviderId = ProviderId(Cow::Borrowed("novita"));
 
     /// Returns all built-in provider IDs
     ///
@@ -100,6 +101,7 @@ impl ProviderId {
             ProviderId::MINIMAX,
             ProviderId::CODEX,
             ProviderId::OPENCODE_ZEN,
+            ProviderId::NOVITA,
         ]
     }
 
@@ -123,6 +125,7 @@ impl ProviderId {
             "io_intelligence" => "IOIntelligence".to_string(),
             "minimax" => "MiniMax".to_string(),
             "codex" => "Codex".to_string(),
+            "novita" => "Novita".to_string(),
             _ => {
                 // For other providers, use UpperCamelCase conversion
                 use convert_case::{Case, Casing};
@@ -165,6 +168,7 @@ impl std::str::FromStr for ProviderId {
             "io_intelligence" => ProviderId::IO_INTELLIGENCE,
             "minimax" => ProviderId::MINIMAX,
             "codex" => ProviderId::CODEX,
+            "novita" => ProviderId::NOVITA,
             // For custom providers, use Cow::Owned to avoid memory leaks
             custom => ProviderId(Cow::Owned(custom.to_string())),
         };
diff --git a/crates/forge_repo/src/provider/provider.json b/crates/forge_repo/src/provider/provider.json
index 1ba9dc8822..4b03b23caa 100644
--- a/crates/forge_repo/src/provider/provider.json
+++ b/crates/forge_repo/src/provider/provider.json
@@ -2383,5 +2383,53 @@
       }
     ],
     "auth_methods": ["api_key"]
+  },
+  {
+    "id": "novita",
+    "api_key_vars": "NOVITA_API_KEY",
+    "url_param_vars": [],
+    "response_type": "OpenAI",
+    "url": "https://api.novita.ai/openai/v1/chat/completions",
+    "models": [
+      {
+        "id": "moonshotai/kimi-k2.5",
+        "name": "Kimi K2.5",
+        "description": "Moonshot AI's flagship MoE model with 262K context, function calling, structured output, reasoning, and vision capabilities",
+        "context_length": 262144,
+        "tools_supported": true,
+        "supports_parallel_tool_calls": true,
+        "supports_reasoning": true,
+        "input_modalities": ["text", "image"]
+      },
+      {
+        "id": "zai-org/glm-5",
+        "name": "GLM-5",
+        "description": "Zhipu AI's flagship MoE model with 202K context, function calling, structured output, and reasoning capabilities",
+        "context_length": 202800,
+        "tools_supported": true,
+        "supports_parallel_tool_calls": true,
+        "supports_reasoning": true,
+        "input_modalities": ["text"]
+      },
+      {
+        "id": "minimax/minimax-m2.5",
+        "name": "MiniMax M2.5",
+        "description": "MiniMax's MoE model with 204K context, function calling, structured output, and reasoning capabilities",
+        "context_length": 204800,
+        "tools_supported": true,
+        "supports_parallel_tool_calls": true,
+        "supports_reasoning": true,
+        "input_modalities": ["text"]
+      },
+      {
+        "id": "qwen/qwen3-embedding-0.6b",
+        "name": "Qwen3 Embedding 0.6B",
+        "description": "Qwen3 embedding model with 1024 dimensions and 8K max input",
+        "context_length": 8192,
+        "tools_supported": false,
+        "input_modalities": ["text"]
+      }
+    ],
+    "auth_methods": ["api_key"]
   }
 ]