feat: add support for GLM 5.2

src/coding_agent_bench/models/__init__.py

@@ -6,6 +6,7 @@
     RedHatAI_Mistral_Small_4_119B_2603_NVFP4,
     RedHatAI_NVIDIA_Nemotron_3_Super_120B_A12B_NVFP4,
     RedHatAI_Qwen3_6_27B_FP8,
+    ZaiOrg_GLM_5_2_FP8,
 )
 
 MODEL_CONFIGS: list[type[ModelConfig]] = [
@@ -15,6 +16,7 @@
     RedHatAI_Mistral_Small_4_119B_2603_NVFP4,
     RedHatAI_NVIDIA_Nemotron_3_Super_120B_A12B_NVFP4,
     RedHatAI_Qwen3_6_27B_FP8,
+    ZaiOrg_GLM_5_2_FP8,
 ]
 
 MODEL_REGISTRY: dict[str, ModelConfig] = {cls.name: cls() for cls in MODEL_CONFIGS}

src/coding_agent_bench/models/configs.py

@@ -14,6 +14,21 @@ class Qwen_Qwen3_8_27B(ModelConfig):
         "--mm-encoder-tp-mode", "data",
     ]
 
+
+class ZaiOrg_GLM_5_2_FP8(ModelConfig):
+
+    name = "zai-org/GLM-5.2-FP8"
+    model_max_len = 131072
+    args = [
+        "--model", "zai-org/GLM-5.2-FP8",
+        "--max-model-len", "131072",
+        "--kv-cache-dtype", "fp8",
+        "--enable-auto-tool-choice",
+        "--tool-call-parser", "glm47",
+        "--reasoning-parser", "glm45",
+    ]
+
+
 class RedHatAI_gemma_4_31B_it_FP8_block(ModelConfig):
 
     name = "RedHatAI/gemma-4-31B-it-FP8-block"

tests/test_models.py

@@ -0,0 +1,17 @@
+from coding_agent_bench.models import get_model_config
+
+
+def test_glm_5_2_fp8_model_config() -> None:
+    config = get_model_config("zai-org/GLM-5.2-FP8")
+
+    assert config.name == "zai-org/GLM-5.2-FP8"
+    assert config.model_max_len == 131072
+    assert config.image == "vllm/vllm-openai:v0.24.0"
+    assert config.args == [
+        "--model", "zai-org/GLM-5.2-FP8",
+        "--max-model-len", "131072",
+        "--kv-cache-dtype", "fp8",
+        "--enable-auto-tool-choice",
+        "--tool-call-parser", "glm47",
+        "--reasoning-parser", "glm45",
+    ]