diff --git a/src/api/providers/openrouter.ts b/src/api/providers/openrouter.ts
index 8d37cdf0a3..ccdde6c378 100644
--- a/src/api/providers/openrouter.ts
+++ b/src/api/providers/openrouter.ts
@@ -97,7 +97,11 @@ export class OpenRouterHandler implements ApiHandler {
 		}
 
 		// Removes messages in the middle when close to context window limit. Should not be applied to models that support prompt caching since it would continuously break the cache.
-		const shouldApplyMiddleOutTransform = !this.getModel().info.supportsPromptCache
+		let shouldApplyMiddleOutTransform = !this.getModel().info.supportsPromptCache
+		// except for deepseek (which we set supportsPromptCache to true for), where because the context window is so small our truncation algo might miss and we should use openrouter's middle-out transform as a fallback to ensure we don't exceed the context window
+		if (this.getModel().id === "deepseek/deepseek-chat") {
+			shouldApplyMiddleOutTransform = true
+		}
 
 		// @ts-ignore-next-line
 		const stream = await this.client.chat.completions.create({
diff --git a/src/core/Cline.ts b/src/core/Cline.ts
index 1bffe8c9d1..9c233b245d 100644
--- a/src/core/Cline.ts
+++ b/src/core/Cline.ts
@@ -52,6 +52,7 @@ import { ClineProvider, GlobalFileNames } from "./webview/ClineProvider"
 import { showSystemNotification } from "../integrations/notifications"
 import { removeInvalidChars } from "../utils/string"
 import { fixModelHtmlEscaping } from "../utils/string"
+import { OpenAiHandler } from "../api/providers/openai"
 
 const cwd =
 	vscode.workspace.workspaceFolders?.map((folder) => folder.uri.fsPath).at(0) ?? path.join(os.homedir(), "Desktop") // may or may not exist but fs checking existence would immediately ask for permission which would be bad UX, need to come up with a better solution
@@ -831,8 +832,26 @@ export class Cline {
 					previousRequest.text,
 				)
 				const totalTokens = (tokensIn || 0) + (tokensOut || 0) + (cacheWrites || 0) + (cacheReads || 0)
-				const contextWindow = this.api.getModel().info.contextWindow || 128_000
-				const maxAllowedSize = Math.max(contextWindow - 40_000, contextWindow * 0.8)
+				let contextWindow = this.api.getModel().info.contextWindow || 128_000
+				// FIXME: hack to get anyone using openai compatible with deepseek to have the proper context window instead of the default 128k. We need a way for the user to specify the context window for models they input through openai compatible
+				if (this.api instanceof OpenAiHandler && this.api.getModel().id.toLowerCase().includes("deepseek")) {
+					contextWindow = 64_000
+				}
+				let maxAllowedSize: number
+				switch (contextWindow) {
+					case 64_000: // deepseek models
+						maxAllowedSize = contextWindow - 25_000
+						break
+					case 128_000: // most models
+						maxAllowedSize = contextWindow - 40_000
+						break
+					case 200_000: // claude models
+						maxAllowedSize = contextWindow - 40_000
+						break
+					default:
+						maxAllowedSize = Math.max(contextWindow - 40_000, contextWindow * 0.8) // for deepseek, 80% of 64k meant only ~10k buffer which was too small and resulted in users getting context window errors.
+				}
+
 				if (totalTokens >= maxAllowedSize) {
 					const truncatedMessages = truncateHalfConversation(this.apiConversationHistory)
 					await this.overwriteApiConversationHistory(truncatedMessages)
diff --git a/src/core/webview/ClineProvider.ts b/src/core/webview/ClineProvider.ts
index 78cc71b460..4081e617dd 100644
--- a/src/core/webview/ClineProvider.ts
+++ b/src/core/webview/ClineProvider.ts
@@ -726,6 +726,11 @@ export class ClineProvider implements vscode.WebviewViewProvider {
 							modelInfo.cacheWritesPrice = 0.3
 							modelInfo.cacheReadsPrice = 0.03
 							break
+						case "deepseek/deepseek-chat":
+							modelInfo.supportsPromptCache = true
+							modelInfo.cacheWritesPrice = 0.14
+							modelInfo.cacheReadsPrice = 0.014
+							break
 					}
 
 					models[rawModel.id] = modelInfo