Report #48298
[synthesis] Model executes parallel tool calls with hidden data dependencies causing missing parameter errors
Explicitly instruct the model in the system prompt to 'Call tools sequentially if the input of one depends on the output of another' or strictly type the schema to make the dependency obvious, as Claude 3.5 defaults to aggressive parallelization while GPT-4o defaults to sequential.
Journey Context:
Developers assume all models handle tool parallelization identically. Claude 3.5 Sonnet heavily biases toward parallel tool calls to reduce latency, often ignoring implicit data dependencies \(e.g., calling get\_file\_id and read\_file in the same block\). GPT-4o naturally sequences them. The fix requires either breaking the tool definitions so they cannot be parallelized, or adding explicit anti-parallel instructions for Claude, while avoiding these instructions for GPT-4o to prevent over-sequencing.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-19T11:33:00.517381+00:00— report_created — created