{
  "recommendations": [
    {
      "model": "glm-5.3-flash:cloud",
      "description": "Fast reasoning for coding and agentic workloads with 1M context and native image understanding",
      "context_length": 1048576,
      "max_output_tokens": 1048576,
      "required_plan": "pro",
      "thinking": {
        "values": [
          "low",
          "high",
          "max"
        ],
        "default": "max"
      }
    },
    {
      "model": "kimi-k3:cloud",
      "description": "Long-horizon agentic reasoning with multimodal tool use",
      "context_length": 1048576,
      "max_output_tokens": 1048576,
      "required_plan": "pro",
      "thinking": {
        "values": [
          false,
          "low",
          "high",
          "max"
        ],
        "default": "max"
      }
    },
    {
      "model": "deepseek-v4.1-flash:cloud",
      "description": "Fast coding and agentic tool use with 1M context and native image understanding",
      "context_length": 1048576,
      "max_output_tokens": 393216,
      "required_plan": "pro",
      "thinking": {
        "values": [
          false,
          "low",
          "high",
          "max"
        ],
        "default": "high"
      }
    },
    {
      "model": "gemma4:31b-cloud",
      "description": "Agentic workflows and multimodal reasoning",
      "context_length": 262144,
      "max_output_tokens": 262144,
      "required_plan": "free",
      "thinking": {
        "values": [
          false,
          true
        ],
        "default": false
      }
    },
    {
      "model": "qwen3.5:4b",
      "description": "Compact multimodal reasoning and tool use",
      "vram_bytes": 6000000000,
      "thinking": {
        "values": [
          false,
          true
        ],
        "default": true
      }
    },
    {
      "model": "gemma4:12b",
      "description": "Agentic workflows and multimodal reasoning",
      "vram_bytes": 11000000000,
      "thinking": {
        "values": [
          false,
          true
        ],
        "default": false
      }
    },
    {
      "model": "qwen3.8:27b",
      "description": "Coding, research, and long-horizon agentic tasks",
      "vram_bytes": 28000000000,
      "thinking": {
        "values": [
          false,
          "low",
          "medium",
          "xhigh"
        ],
        "default": "xhigh"
      }
    }
  ]
}