Unverified50% confidenceEventExact time
OpenAI is secretly testing a frontier experimental model codenamed Bell internally; leaked information claims its capability metrics surpass GPT-6 Astra.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/1/2026
First Seen
Valid until: 12/30/2026
Sources
Related Entities
Related Claims
UnverifiedOpenAI used Semi-Analysis's public benchmark InferenceX to test Jalapeño on three open models: GPT-OSS 120B, DeepSeek R1 670B, and Kimi K2 1T.72% similarVerifiedOpenAI released the next-generation large model GPT-6 Astra.70% similarUnverifiedAccording to leaked information, an internally developed OpenAI model codenamed Bell is a pre-trained model with more than 10 trillion parameters.70% similarUnverifiedA previously undisclosed model entry appeared in OpenAI's help center, listed as a supported mainline model built on the Daybreak core architecture and carrying over the GPT-6 safety protection system, and was subsequently deleted by OpenAI.68% similarUnverified匿名模型gpt2-chatbot曾在Arena上引发广泛猜测，最终被证实与OpenAI相关68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/963821API
curl https://kongchang.com/api/v1/knowledge/claims/963821MCP
get_claim(id=963821)