A list of AI models that fit within the memory constraints of a specific GPU for a given context length, including reasoning and recommendations.
The agent uses the 'find_fit' tool from the AI Compute Radar MCP server. It requires the user to specify the target hardware (e.g., 'rtx-4090') and the desired context length. The agent sends this request to the server, which returns a verdict including measured weights, computed context cache, overhead, and a plain-language explanation of whether the model fits. The agent should verify the 'meta' block to ensure the data source is 'live' and check the collection time to ensure freshness.
- Category: Software development
- Effect: Read-only
- MCP servers: AI Compute Radar MCP (https://koskamo.com/explore/ai-compute-radar-mcp)
- Source: Developers — JSON API and MCP server — AI Compute Radar (https://aicomputeradar.dev/developers)
- Source quote: "Which tracked models fit an RTX 4090 at 16K context: curl "https://aicomputeradar.dev/api/v1/fit?hardware=rtx-4090&context=16384""
- Playbook page: https://koskamo.com/explore/use-cases/find-ai-models-fitting-specific-hardware
Prompt: Find which AI models fit on an RTX 4090 with a 16K context window. Return the fit verdicts, including the reasons for any failures and the best recommendation.