nvidia/gpt-oss-120b-Eagle3-long-context

Different architecture vs openai/gpt-oss-120b. Different architecture: only 1 of 687/12 tensors match by name and shape.

1 other repos hold byte-for-byte identical weight files.