{ "model_id": "glm-5.2-visual-runtime", "kind": "training_free_multimodal_gateway", "checkpoint_repo": null, "reasoning_model": { "provider": "vllm_openai_compatible", "recommended_checkpoint": "zai-org/GLM-5.2-FP8", "served_model_name": "glm-5.2", "minimum_vllm_version": "0.23.0", "recommended_image": "vllm/vllm-openai:glm52" }, "vision_model": { "provider": "local_vllm_omni_openai_compatible", "recommended_checkpoint": "Qwen/Qwen3-Omni-30B-A3B-Instruct", "served_model_name": "qwen3-omni", "endpoint": "http://vision-vllm:8000/v1" }, "alternate_reasoning_model": { "provider": "vllm_openai_compatible", "recommended_checkpoint": "Qwen/Qwen3.6-27B", "served_model_name": "qwen3.6-27b", "minimum_vllm_version": "0.19.0" }, "ocr_model": { "provider": "local_container", "implementation": "one_click/ocr", "endpoint": "http://ocr:8080/ocr" }, "visual_runtime": { "space_repo": "wassemgtk/glm-5-2-visual-runtime-space", "public_model_id": "glm-5.2-visual-runtime", "default_mode": "local", "local_mode_includes": [ "gateway", "zai-org/GLM-5.2-FP8 via vLLM", "Qwen/Qwen3-Omni-30B-A3B-Instruct via vLLM-Omni", "local OCR container", "PostgreSQL", "MinIO" ], "cloud_mode_requires": [ "GLM_BASE_URL", "VISION_BASE_URL", "DATABASE_URL", "OBJECT_STORE_ENDPOINT" ] }, "vllm_arguments": { "model": "zai-org/GLM-5.2-FP8", "served_model_name": "glm-5.2", "kv_cache_dtype": "fp8", "tensor_parallel_size": 8, "tool_call_parser": "glm47", "reasoning_parser": "glm45", "enable_auto_tool_choice": true, "speculative_config": { "method": "mtp", "num_speculative_tokens": 5 } }, "openai_compatible_endpoints": [ "/v1/chat/completions", "/v1/responses" ], "notes": [ "This repo intentionally does not include GLM-5.2 weights.", "The one-click profile pulls GLM-5.2-FP8 and Qwen3-Omni weights from Hugging Face at deployment time unless models/ has been materialized.", "Run scripts/materialize_weights.py and upload models/ if you want all checkpoint weights physically stored in this repo.", "No hosted model provider is required in VISUAL_RUNTIME_MODE=local." ] }