Updated: October 2026
The id8 Best LLM for Tool Calling guide ranks 164 models. Tool calling needs precise, format-correct actions; models are judged on agentic coding and instruction following. Right now GPT-6 Luna, Qwen3.8 Max and GPT-6.1 Sol lead (graded on SWE-bench Verified, Aider Polyglot and IFEval plus an overall task grade). Best open-weight: DeepSeek V4.1 Flash. Prices and scores are re-checked daily.
Right now GPT-6 Luna, Qwen3.8 Max and GPT-6.1 Sol lead (graded on SWE-bench Verified, Aider Polyglot and IFEval plus an overall task grade). Best open-weight: DeepSeek V4.1 Flash.
Tool calling, also called function calling, is when a model returns a structured request — a function name and arguments — instead of plain text. Your code runs the function and sends the result back, and the model continues from there.
DeepSeek V4.1 Flash, Kimi K3 and Qwen 3 235B-A22B lead the open-weight models for tool calling right now. Validate every call against your schema; smaller models are more likely to return malformed arguments.
The common causes are an invented function name, arguments that do not match the schema, and calling tools in the wrong order. Clear function descriptions and strict schema validation remove most of these.
Related Tools
Related Guides