Tuna-2

Summary

Tuna-2 is a pixel-space unified multimodal model that discards pretrained vision encoder modules.

Role In The Wiki

Tuna-2 is the strongest current counterpoint to semantic-encoder-first multimodal design in the corpus.

Evidence

Relation To Foundation TSFM Agenda

Use the source-level agenda mapping in tuna-2-2026 rather than duplicating verdict rows here.

At the entity level, Tuna-2 is the strongest current counterpoint to semantic-encoder-first multimodal design in the corpus. This page should stay as the object card; source pages carry slot-level verdicts, evidence, and missing pieces.