Go to issue

Labels Milestones New Issue

1 Open 115 Closed

Label

Show archived labels Use alt + click/enter to exclude labels

All labels No label

cannot reproduce

bug

critical

duplicate

enhancement

help wanted

idea

invalid

question

wontfix

Milestone

All milestones No milestones

Closed milestones

Mojo GPU Backend Transition

Project

All projects No project

Author

All users

Assignee

Assigned to nobody Assigned to anybody

hugo

Sort

Newest Oldest Most recently updated Least recently updated Most commented Least commented Nearest due date Farthest due date

1 Open 115 Closed

Open

Label

Clear labels

cannot reproduce

bug

critical

duplicate

enhancement

help wanted

idea

invalid

question

wontfix

Milestone

No milestone

Closed milestones

Mojo GPU Backend Transition

Projects

Clear projects

Assignee

Clear assignees

hugo

Match Qwen direct-mode reasoning history rendering bug

#117

by hugo was closed 2026-10-04 15:38:49 +00:00

1

Apply the Qwen tokenizer NFC normalization contract bug

#116

by hugo was closed 2026-10-04 15:14:22 +00:00

1

Quote numeric-looking revisions in the Dev Brain Git fixture bug

#115

by hugo was closed 2026-10-04 15:01:06 +00:00

1

Preserve Qwen prefill cached-key zero sign bug

#114

by hugo was closed 2026-10-04 14:56:06 +00:00

1

Preserve MTPLX head-chunked Qwen verify SDPA execution bug

#113

by hugo was closed 2026-10-04 14:45:51 +00:00

1

Port MTPLX row-map SwiGLU expert gather for Qwen prefill MoE bug

#112

by hugo was closed 2026-10-04 14:25:29 +00:00

1

Tokenize all Unicode number categories with the Qwen3 contract bug

#111

by hugo was closed 2026-10-04 13:37:37 +00:00

3

Prevent uninitialized expert outputs for large unaligned M5 prefill chunks bug

#110

by hugo was closed 2026-10-04 13:18:52 +00:00

6

Preserve compiled BF16 attention-gate math in MTP verification bug

#109

by hugo was closed 2026-10-04 12:50:58 +00:00

3

Apply M-RoPE row by row across the vision position-table boundary bug

#108

by hugo was closed 2026-10-04 12:11:47 +00:00

3

Preserve sparse QSA and image-aware indexer positions for vision turns bug

#107

by hugo was closed 2026-10-04 12:06:01 +00:00

3

Commit the final MTP primary before banking warm session state bug

#106

by hugo was closed 2026-10-04 11:42:37 +00:00

3

Reject invalid Qwen logits instead of emitting fabricated tokens bug

#105

by hugo was closed 2026-10-04 11:29:45 +00:00

3

Use full-vocabulary top-p law in Qwen AR pipeline bug

#104

by hugo was closed 2026-10-04 10:50:18 +00:00

5

Analyse Qwen3.8 Flash Next changes in MTPLX and create confirmed bug issues criticalenhancement

#103

by hugo was closed 2026-10-04 06:28:57 +00:00

4

opening GLM 5.3 flash session when qwen is selected gives an error cannot reproducebug

#102

by hugo was closed 2026-10-04 10:27:11 +00:00

Support switching model via direct dropdown on model name in chat while still empty enhancement

#101

by hugo was closed 2026-10-02 08:35:38 +00:00

1

support deepseek v4 flash vision experimental enhancement

#100

by hugo was closed 2026-10-04 18:33:20 +00:00

1

Optimize and benchmark Qwen3.8 Flash Next on Apple silicon enhancement

#99

by hugo was closed 2026-09-04 21:28:16 +00:00

1

Add Qwen3.8 Flash Next native MTP speculation enhancement

#98

by hugo was closed 2026-09-03 20:39:48 +00:00

1