Block a user
check rustpython to see if we can go back to python as scripting language
add support for a separate vision model (small) to add eyes
add support for a separate vision model (small) to add eyes
not needed anymore due to fast vision capable models and ds4 flash exp
ssd streaming seems to not show enough in the graph
support deepseek v4 flash vision experimental
Optimize and benchmark Qwen3.8 Flash Next on Apple silicon
Optimize and benchmark Qwen3.8 Flash Next on Apple silicon
Implemented and delivered in commit bd63538 (pushed to origin/main).
The Qwen3.8 Flash Next path now follows the pinned MTPLX/MLX execution contract end to end: exact affine QMV/QMM and BF16…
Add Qwen3.8 Flash Next native MTP speculation
Add Qwen3.8 Flash Next native MTP speculation
Implemented in commit bf82df7 (Add native Qwen MTP speculation).
Implementation:
- Added the complete Rust Qwen3.8 one-layer MTP graph: four-stream hidden and token projections, full QSA…
Implement exact Qwen Sparse Attention and long-context state
Implement exact Qwen Sparse Attention and long-context state
Implement exact Qwen Sparse Attention and long-context state
Implement exact Qwen Sparse Attention and long-context state
Implemented and verified in dda70c3 (Implement exact Qwen sparse attention).
Implementation:
- Added the official Qwen indexer graph in Rust: one affine 640-wide projection split into four…
Stream Qwen3.8 Flash Next PLE from mapped storage
Stream Qwen3.8 Flash Next PLE from mapped storage
Implemented fully in commit 87ccf67 (pushed to origin/main).
Implementation:
- Added the official Qwen3.8 Flash Next 2/3-gram PLE hash contract in Rust, including the exact splitmix multipliers,…