{"slug":"glm-5","id":"glm-5","type":"model","title":"GLM-5","description":"Large MoE model with strongest coding benchmark among open models (77.8% SWE-bench). 50.4% on Humanity's Last Exam. MIT license with no usage restrictions.","last_updated":"2026-04-10","last_verified":null,"verification_status":"unverified","markdown_url":"/content/models/glm-5.md","html_url":"/models/glm-5","api_url":"/api/v1/models/glm-5.json","content_hash":"a76745362d5dfd180eaf3a17bf8ca609a5de1b9e08f6c03d6f60d79b7c16184a","sha256":"a76745362d5dfd180eaf3a17bf8ca609a5de1b9e08f6c03d6f60d79b7c16184a","provider":"Zhipu AI","pricing":{"input":"Free (self-hosted)","output":"Free (self-hosted)","free":true,"note":"Also via Zhipu API"},"benchmarks":{"reasoning":90,"coding":93,"math":88,"writing":82,"multilingual":83,"speed":70},"tags":["zhipu ai","open-source","text"],"website":"https://www.zhipuai.cn","release_date":"2026-02","relationships":{"links":[],"related":[{"id":"cohere-tiny-aya","title":"Cohere Tiny Aya 3.35B","type":"model","html_url":"/models/cohere-tiny-aya","markdown_url":"/content/models/cohere-tiny-aya.md","shared_tags":["open-source","text"],"score":4},{"id":"command-r-plus","title":"Command R+","type":"model","html_url":"/models/command-r-plus","markdown_url":"/content/models/command-r-plus.md","shared_tags":["open-source","text"],"score":4},{"id":"deepseek-r1","title":"DeepSeek R1","type":"model","html_url":"/models/deepseek-r1","markdown_url":"/content/models/deepseek-r1.md","shared_tags":["open-source","text"],"score":4},{"id":"deepseek-v3.2","title":"DeepSeek V3.2","type":"model","html_url":"/models/deepseek-v3.2","markdown_url":"/content/models/deepseek-v3.2.md","shared_tags":["open-source","text"],"score":4},{"id":"falcon-3","title":"Falcon 3","type":"model","html_url":"/models/falcon-3","markdown_url":"/content/models/falcon-3.md","shared_tags":["open-source","text"],"score":4},{"id":"gemma-3","title":"Gemma 3","type":"model","html_url":"/models/gemma-3","markdown_url":"/content/models/gemma-3.md","shared_tags":["open-source","text"],"score":4}],"explicit":{}},"metadata":{"title":"GLM-5","type":"model","id":"glm-5","provider":"Zhipu AI","model_type":"open-source","release_date":"2026-02","description":"Large MoE model with strongest coding benchmark among open models (77.8% SWE-bench). 50.4% on Humanity's Last Exam. MIT license with no usage restrictions.","last_updated":"2026-04-10","context_window":"128K tokens","website":"https://www.zhipuai.cn","license":"MIT","modality":["text"],"tags":["zhipu ai","open-source","text"],"pricing":{"input":"Free (self-hosted)","output":"Free (self-hosted)","free":true,"note":"Also via Zhipu API"},"benchmarks":{"reasoning":90,"coding":93,"math":88,"writing":82,"multilingual":83,"speed":70},"parameters":"744B total (40B active)","hardware_requirements":"8x A100 80GB (FP16); 2x A100 with Q4 quantization","best_for":["Code generation","Complex reasoning","Enterprise deployment","Research"]},"content_text":"# GLM-5\n\nThe best open-source coding model, full stop. GLM-5's 77.8% on SWE-bench beats every other open-weight model and most proprietary ones, while its 50.4% on Humanity's Last Exam puts it in rare company for general reasoning. Zhipu AI came out of nowhere for Western audiences, but these numbers speak for themselves.\n\nThe coding benchmark score of 93/100 is not a fluke -- it translates directly to real-world code generation tasks. Reasoning (90) and math (88) are similarly strong. The weak spot is speed at 70/100, which is the tax you pay for a 744B MoE architecture even with only 40B parameters active per token. Writing at 82 is competent but not the reason you pick this model.\n\nSelf-hosting under MIT with zero usage restrictions is the dream license for enterprise. The hardware cost is real though: 8x A100 80GB for FP16, or 2x A100 if you quantize to Q4. That puts it firmly in \"serious infrastructure\" territory, not hobbyist-friendly. The Zhipu API exists if you want to skip the hardware bill.\n\nThe main gap is ecosystem. Western tooling integration is still thin compared to Llama or Qwen, and community fine-tunes are sparse. If you need a coding powerhouse and can handle the infrastructure, nothing open-source touches it.\n\n**When to pick something else:** For a more balanced open model with better multilingual support and a larger community, Qwen 3.5 is the safer choice. For coding on consumer hardware, Nemotron-Cascade 2 delivers remarkable results at a fraction of the compute.","content_length":2383,"generated_at":"2026-04-24"}