hugo - Hugo's house of code horror

hugo commented on issue hugo/DS4Server#93

2026-09-03 17:16:24 +00:00

Prepare exact Qwen3.8 Flash Next Bare Speed artifacts

Implemented and pushed as commit 4c83ac0.

The repository now contains a Rust-only qwen38-artifacts tool plus an immutable manifest and complete 2,527-tensor inventory for Youssofal/Qwen3.8-Flash-N…

hugo pushed to main at hugo/DS4Server

2026-09-03 17:16:05 +00:00

4c83ac0360 Prepare Qwen3.8 artifact set

hugo pushed to main at hugo/DS4Server

2026-09-01 20:52:21 +00:00

9556b26c55 Bind sessions to their selected model

hugo pushed to main at hugo/DS4Server

2026-09-01 20:38:34 +00:00

35f306bf2a Lock tool sessions to their model

hugo pushed to main at hugo/DS4Server

2026-09-01 19:03:17 +00:00

d40e86e5ef correct thinking modes for each model

hugo opened issue hugo/DS4Server#98

2026-09-01 18:24:31 +00:00

Add Qwen3.8 Flash Next native MTP speculation

hugo opened issue hugo/DS4Server#99

2026-09-01 18:24:31 +00:00

Optimize and benchmark Qwen3.8 Flash Next on Apple silicon

hugo opened issue hugo/DS4Server#95

2026-09-01 18:24:30 +00:00

Implement the Qwen3.8 Flash Next Rust and Metal text core

hugo opened issue hugo/DS4Server#96

2026-09-01 18:24:30 +00:00

Stream Qwen3.8 Flash Next PLE from mapped storage

hugo opened issue hugo/DS4Server#97

2026-09-01 18:24:30 +00:00

Implement exact Qwen Sparse Attention and long-context state

hugo opened issue hugo/DS4Server#94

2026-09-01 18:24:15 +00:00

Define Qwen3.8 Flash Next artifacts and model contract

hugo pushed to main at hugo/DS4Server

2026-09-01 18:09:37 +00:00

ee515ee824 Load GLM vision weights on demand

hugo pushed to main at hugo/DS4Server

2026-09-01 17:28:59 +00:00

46d6a976a5 Add GLM 5.3 Flash support

hugo opened issue hugo/DS4Server#93

2026-09-01 11:53:12 +00:00

support qwen 3.8 flash next

hugo commented on issue hugo/DS4Server#87

2026-08-31 15:31:51 +00:00

Implement the complete DeepSeek Mojo-Metal backend with performance parity