From 197937aab0ae219fad5268fe9c8fa228728385f1 Mon Sep 17 00:00:00 2001
From: Concedo <39025047+LostRuins@users.noreply.github.com>
Date: Tue, 16 Jun 2026 01:03:24 +0800
Subject: [PATCH] cohere think budget

---
 gpttype_adapter.cpp | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/gpttype_adapter.cpp b/gpttype_adapter.cpp
index d475dc56f..97dab95e0 100644
--- a/gpttype_adapter.cpp
+++ b/gpttype_adapter.cpp
@@ -5429,6 +5429,11 @@ generation_outputs gpttype_generate(const generation_inputs inputs)
                 end = "</seed:think>";
                 budget_exceeded = "\n(Reasoning budget exceeded)\n<seed:cot_budget_reflect>The current thinking budget is 0, so I will directly start answering the question.</seed:cot_budget_reflect>\nTime to respond now.\n</seed:think>";
                 break;
+            case llm_arch::LLM_ARCH_COHERE2MOE:
+                start = "<|START_THINKING|>";
+                end = "<|END_THINKING|>";
+                 budget_exceeded = "\n(Reasoning budget exceeded)\nTime to respond now.\n<|END_THINKING|>";
+                break;
             default:
                 break;
         }