From 2d8bce7ccb279c20f9eb5ebc075bf431a54d875b Mon Sep 17 00:00:00 2001
From: Sergey Kozyrenko <sirozha_blr@tut.by>
Date: Fri, 26 Jun 2026 03:11:20 +0700
Subject: [PATCH] feat(providers): add glm-5.2, kimi-k2.7-code, qwen3.7-plus to
 model catalogs

- glm/models.yml: add glm-5.2 (only GLM model supporting reasoning_effort=high/max)
- kimi/models.yml: add kimi-k2.7-code (code-specialist above k2.6)
- qwen/models.yml: add qwen3.7-plus (cost-efficient tier below qwen3.7-max)

Prices for kimi-k2.7-code and qwen3.7-plus are estimated from adjacent
catalog entries and should be verified against official pricing pages.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
---
 backend/pkg/providers/glm/models.yml  |  8 ++++++++
 backend/pkg/providers/kimi/models.yml |  9 +++++++++
 backend/pkg/providers/qwen/models.yml | 11 ++++++++++-
 3 files changed, 27 insertions(+), 1 deletion(-)

diff --git a/backend/pkg/providers/glm/models.yml b/backend/pkg/providers/glm/models.yml
index 76626eab..3f48963c 100644
--- a/backend/pkg/providers/glm/models.yml
+++ b/backend/pkg/providers/glm/models.yml
@@ -1,4 +1,12 @@
 # GLM-5.x Series - Latest generation (200K context, 128K max output, hybrid thinking)
+- name: glm-5.2
+  description: GLM-5.2 - Newest flagship (improves on GLM-5.1), 200K context, hybrid thinking; the only GLM model that honors reasoning_effort (high/max).
+  thinking: true
+  price:
+    input: 1.40
+    output: 4.40
+    cache_read: 0.26
+
 - name: glm-5.1
   description: GLM-5.1 - Latest flagship designed for long-horizon tasks (8h sustained autonomous execution), Claude Opus 4.6-aligned coding, 200K context, hybrid thinking. Best for planning, mentor, and complex agentic engineering
   thinking: true
diff --git a/backend/pkg/providers/kimi/models.yml b/backend/pkg/providers/kimi/models.yml
index b18d33e3..62f0ba9a 100644
--- a/backend/pkg/providers/kimi/models.yml
+++ b/backend/pkg/providers/kimi/models.yml
@@ -1,3 +1,12 @@
+# Kimi K2.7 Code - Code-specialized flagship model
+- name: kimi-k2.7-code
+  description: Kimi K2.7 Code - Code-specialist variant of K2.7 with native architecture. Stronger code generation and long-term reasoning than K2.6; supports thinking/non-thinking modes, 256k context, automatic context caching, ToolCalls, JSON Mode
+  thinking: true
+  price:
+    input: 1.20
+    output: 5.00
+    cache_read: 0.20
+
 # Kimi K2.6 - Latest flagship multimodal model with native architecture
 - name: kimi-k2.6
   description: Kimi K2.6 - Latest and most intelligent multimodal model with native architecture, stronger long-term code writing, improved instruction compliance and self-correction. Supports text/image/video input, thinking/non-thinking modes, 256k context, automatic context caching, ToolCalls, JSON Mode, internet search
diff --git a/backend/pkg/providers/qwen/models.yml b/backend/pkg/providers/qwen/models.yml
index 6771f9d7..3162cf59 100644
--- a/backend/pkg/providers/qwen/models.yml
+++ b/backend/pkg/providers/qwen/models.yml
@@ -1,4 +1,13 @@
-# Qwen3.7 Series - Latest flagship Max (May 2026)
+# Qwen3.7 Series - Latest generation (May 2026)
+- name: qwen3.7-plus
+  description: Qwen3.7 Plus - Cost-efficient tier of the Qwen3.7 generation for the agent-centric era. Better cost-to-intelligence ratio than qwen3.7-max; supports hybrid thinking, long-horizon autonomous execution, and complex security reasoning workflows
+  thinking: true
+  release_date: 2026-05-21
+  price:
+    input: 1.0
+    output: 3.5
+    cache_read: 0.1
+
 - name: qwen3.7-max
   description: Qwen3.7 Max - Next-generation flagship designed for the agent-centric era with excellence in programming, productivity tasks, and long-term autonomous execution. Pure-text interface optimal for sophisticated penetration testing automation and complex agentic security workflows
   thinking: true