{
  "recommendations": [
    {
      "model": "minimax-m3:cloud",
      "description": "State-of-the-art coding \u0026 agent capabilities with multimodal reasoning and support for up to 1M context",
      "context_length": 524288,
      "max_output_tokens": 131072,
      "required_plan": "free"
    },
    {
      "model": "kimi-k2.6:cloud",
      "description": "State-of-the-art coding, long-horizon execution, and multimodal agent swarm capability",
      "context_length": 262144,
      "max_output_tokens": 262144,
      "required_plan": "pro"
    },
    {
      "model": "glm-5.1:cloud",
      "description": "Long-horizon agentic engineering with autonomous execution and sustained iteration",
      "context_length": 202752,
      "max_output_tokens": 131072,
      "required_plan": "pro"
    },
    {
      "model": "qwen3.5:cloud",
      "description": "Reasoning, coding, and agentic tool use with vision",
      "context_length": 262144,
      "max_output_tokens": 32768,
      "required_plan": "pro"
    },
    {
      "model": "nemotron-3-super:cloud",
      "description": "Reasoning, coding, and agentic tool use",
      "context_length": 262144,
      "max_output_tokens": 65536,
      "required_plan": "free"
    },
    {
      "model": "gemma4:31b-cloud",
      "description": "Agentic workflows and multimodal reasoning",
      "context_length": 262144,
      "max_output_tokens": 262144,
      "required_plan": "free"
    },
    {
      "model": "gemma4",
      "description": "Agentic workflows and multimodal reasoning",
      "vram_bytes": 12000000000
    },
    {
      "model": "qwen3.6",
      "description": "Agentic software engineering model with multimodal capabilities",
      "vram_bytes": 28000000000
    }
  ]
}