diff --git a/.changeset/drop-max-tokens-gpt5-openai-compatible.md b/.changeset/drop-max-tokens-gpt5-openai-compatible.md
new file mode 100644
index 00000000000..cdf54e8d48d
--- /dev/null
+++ b/.changeset/drop-max-tokens-gpt5-openai-compatible.md
@@ -0,0 +1,5 @@
+---
+"@kilocode/cli": patch
+---
+
+Fix gpt-5 models failing with `Unsupported parameter: max_tokens` when accessed through custom OpenAI-compatible providers such as LiteLLM.
diff --git a/packages/opencode/src/session/llm.ts b/packages/opencode/src/session/llm.ts
index 39164727d12..7cdb5e85c05 100644
--- a/packages/opencode/src/session/llm.ts
+++ b/packages/opencode/src/session/llm.ts
@@ -186,7 +186,14 @@ const live: Layer.Layer<
             : undefined,
           topP: input.agent.topP ?? ProviderTransform.topP(input.model),
           topK: ProviderTransform.topK(input.model),
-          maxOutputTokens: ProviderTransform.maxOutputTokens(input.model),
+          // kilocode_change start - gpt-5 via @ai-sdk/openai-compatible proxies (e.g. LiteLLM)
+          // rejects `max_tokens`; OpenAI requires `max_completion_tokens` and the compatible
+          // SDK cannot rename the field, so drop the cap and let the upstream default apply.
+          maxOutputTokens:
+            input.model.api.npm === "@ai-sdk/openai-compatible" && input.model.api.id.toLowerCase().includes("gpt-5")
+              ? undefined
+              : ProviderTransform.maxOutputTokens(input.model),
+          // kilocode_change end
           options,
         },
       )