Unverified50% confidenceTradeoffExact time
开放权重模型的工具调用格式往往不如闭源模型稳定，闭源模型通过大量RLHF、后训练对齐和运行时约束（如JSON Mode、Structured Output）确保工具调用格式精确性
1
Sources
50%
Confidence
Long-term
Relevance
8/26/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAchieving compatibility with the OpenAI format requires precise alignment of SSE streaming output format, JSON Schema definitions for Function Calling, and error code semantic mapping, enabling proxy tools such as LiteLLM to achieve transparent compatibility.74% similarUnverifiedThe vast majority of so-called open-source models only release model weights, while keeping training code, data processing pipelines, and training configurations in a black box.73% similarVerifiedOpenAI's Structured Output is based on Constrained Sampling technology, which allows only tokens conforming to the current JSON Schema constraints to be sampled at each decoding step, thereby mathematically guaranteeing the validity of the output format.72% similarUnverified当Thinking模式和工具调用同时启用时，OpenAI格式会强制返回推理过程导致兼容性问题，而Anthropic格式能确保两种能力稳定共存71% similarUnverifiedMainstream model APIs from OpenAI, Anthropic, and others all support passing in a JSON Schema at call time to constrain output format71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/802894API
curl https://kongchang.com/api/v1/knowledge/claims/802894MCP
get_claim(id=802894)