{"slug": "opencode-data-real-world-ai-model-usage-cache-ratios-and-costs", "title": "OpenCode Data: Real-world AI model usage, cache ratios, and costs", "summary": "OpenCode released data on real-world AI model usage, cache ratios, and costs, showing DeepSeek's v4-flash leading with 8.2T tokens, followed by DeepSeek v4-pro and Zhipu's glm-5.2. The data reveals shifting model adoption and cost implications for developers.", "body_md": "[Model Data](#overview)\n\nSee which models are winning real usage, how the mix is shifting, and what that means for cost.\n\nTop Models. Usage of models across OpenCode Go.\n\n[Top Models.](#top-models)\n\n[04](/data/minimax/minimax-m3)\n\n**minimax-m3** 960B\n\nMiniMax-61%\n\n[05](/data/xiaomi/mimo-v2-5)\n\n**mimo-v2.5** 827B\n\nXiaomi-4%\n\n[06](/data/moonshot/kimi-k2-7-code)\n\n**kimi-k2.7-code** 391B\n\nMoonshot-17%\n\n[07](/data/xiaomi/mimo-v2-5-pro)\n\n**mimo-v2.5-pro** 273B\n\nXiaomi-5%\n\n[08](/data/qwen/qwen3-7-plus)\n\n**qwen3.7-plus** 247B\n\nQwen+15%\n\n[09](/data/moonshot/kimi-k2-6)\n\n**kimi-k2.6** 118B\n\nMoonshot-15%\n\n[10](/data/zhipu/glm-5-1)\n\n**glm-5.1** 59B\n\nZhipu-20%\n\n[11](/data/minimax/minimax-m2-7)\n\n**minimax-m2.7** 54B\n\nMiniMax-14%\n\n[12](/data/qwen/qwen3-7-max)\n\n**qwen3.7-max** 54B\n\nQwen-4%\n\n[13](/data/qwen/qwen3-6-plus)\n\n**qwen3.6-plus** 39B\n\nQwen-2%\n\n[14](/data/moonshot/kimi-k2-5)\n\n**kimi-k2.5** 6B\n\nMoonshot-19%\n\n[15](/data/zhipu/glm-5)\n\n**glm-5** 5B\n\nZhipu-14%\n\n[16](/data/minimax/minimax-m2-5)\n\n**minimax-m2.5** 5B\n\nMiniMax-7%\n\n[17](/data/qwen/qwen3-5-plus)\n\n**qwen3.5-plus** 2B\n\nQwen-45%\n\n[18](/data/tencent/hy3-preview)\n\n**hy3-preview** 0B\n\nTencent-58%\n\n[01](/data/deepseek/deepseek-v4-flash)\n\n**deepseek-v4-flash** 8.2T\n\nDeepSeek-20%\n\n[02](/data/deepseek/deepseek-v4-pro)\n\n**deepseek-v4-pro** 2.8T\n\nDeepSeek-27%\n\n[03](/data/zhipu/glm-5-2)\n\n**glm-5.2** 1.3T\n\nZhipu-17%\n\n[04](/data/minimax/minimax-m3)\n\n**minimax-m3** 960B\n\nMiniMax-61%\n\n[05](/data/xiaomi/mimo-v2-5)\n\n**mimo-v2.5** 827B\n\nXiaomi-4%\n\n[06](/data/moonshot/kimi-k2-7-code)\n\n**kimi-k2.7-code** 391B\n\nMoonshot-17%\n\n[07](/data/xiaomi/mimo-v2-5-pro)\n\n**mimo-v2.5-pro** 273B\n\nXiaomi-5%\n\n[08](/data/qwen/qwen3-7-plus)\n\n**qwen3.7-plus** 247B\n\nQwen+15%\n\n[09](/data/moonshot/kimi-k2-6)\n\n**kimi-k2.6** 118B\n\nMoonshot-15%\n\n[10](/data/zhipu/glm-5-1)\n\n**glm-5.1** 59B\n\nZhipu-20%\n\n[11](/data/minimax/minimax-m2-7)\n\n**minimax-m2.7** 54B\n\nMiniMax-14%\n\n[12](/data/qwen/qwen3-7-max)\n\n**qwen3.7-max** 54B\n\nQwen-4%\n\n[13](/data/qwen/qwen3-6-plus)\n\n**qwen3.6-plus** 39B\n\nQwen-2%\n\n[14](/data/moonshot/kimi-k2-5)\n\n**kimi-k2.5** 6B\n\nMoonshot-19%\n\n[15](/data/zhipu/glm-5)\n\n**glm-5** 5B\n\nZhipu-14%\n\n[16](/data/minimax/minimax-m2-5)\n\n**minimax-m2.5** 5B\n\nMiniMax-7%\n\n[17](/data/qwen/qwen3-5-plus)\n\n**qwen3.5-plus** 2B\n\nQwen-45%\n\n[18](/data/tencent/hy3-preview)\n\n**hy3-preview** 0B\n\nTencent-58%\n\n[LEARN MORE](#top-models)\n\n**TOP MODELS****▸**** Unique Users.** Daily unique OpenCode Go users by model.", "url": "https://wpnews.pro/news/opencode-data-real-world-ai-model-usage-cache-ratios-and-costs", "canonical_source": "https://opencode.ai/data/", "published_at": "2026-07-08 14:32:00+00:00", "updated_at": "2026-07-08 14:43:08.588289+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-products", "ai-infrastructure", "ai-tools"], "entities": ["OpenCode", "DeepSeek", "Zhipu", "MiniMax", "Moonshot", "Qwen", "Xiaomi", "Tencent"], "alternates": {"html": "https://wpnews.pro/news/opencode-data-real-world-ai-model-usage-cache-ratios-and-costs", "markdown": "https://wpnews.pro/news/opencode-data-real-world-ai-model-usage-cache-ratios-and-costs.md", "text": "https://wpnews.pro/news/opencode-data-real-world-ai-model-usage-cache-ratios-and-costs.txt", "jsonld": "https://wpnews.pro/news/opencode-data-real-world-ai-model-usage-cache-ratios-and-costs.jsonld"}}