Skip to content

Pull requests: ml-explore/mlx-lm

New pull request New
89 Open 570 Closed
89 Open 570 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

chore: Use json-repair for json_tools tool-call parsing
#1071 opened Mar 29, 2026 by isaac-cf-wong Loading…
Add TurboQuant KV cache compression (3-bit, 4.6x)
#1067 opened Mar 28, 2026 by arozanov Loading…
4 of 6 tasks
12
Fix gated delta kernel precision
#1066 opened Mar 27, 2026 by kernelpool • Draft
1
[Experimental] Add TurboQuantKVCache: PolarQuant KV cache compression at 2-4 bits
#1059 opened Mar 26, 2026 by rachittshah Loading…
1
Add LongCat Next
#1057 opened Mar 26, 2026 by kernelpool Loading…
lfm2: strip lm_head.weight for tied embeddings
#1055 opened Mar 25, 2026 by ykhrustalev • Draft
2
Fix IndexError in CacheOrder.pop() on empty cache
#1054 opened Mar 25, 2026 by lyonsno • Draft
2 tasks done
Fix BatchRotatingKVCache.merge() OOB when prompt exceeds max_size
#1052 opened Mar 25, 2026 by LxYuan0420 Loading…
fix tokenizer regex issue with Mistral-based models
#1049 opened Mar 23, 2026 by amanning3390 Loading…
Fix zero prompt-cache reuse for thinking models in multi-turn chat
#1042 opened Mar 22, 2026 by lyonsno Loading…
10
skip xtc sampling when threshold is zero
#1040 opened Mar 22, 2026 by mm65x • Draft
1
Fix prompt cache leak between conversations
#1039 opened Mar 22, 2026 by kernelpool Loading…
1
feat: configurable KVCache step size and pre-allocation
#1038 opened Mar 22, 2026 by Thump604 Loading…
5 tasks
1
Add Mistral Small 4 (119B MoE) support via mistral4.py
#1037 opened Mar 21, 2026 by ProducerGuy Loading…
5 tasks done
Handle Metal OOM gracefully in mlx_lm.server with structured errors
#1034 opened Mar 21, 2026 by Aristide021 Loading…
1
1
feat: Drop unsupported weights by default during model load
#1033 opened Mar 21, 2026 by spicyneuron Loading…
1
Fuse gate/up expert projections in SwitchGLU
#1032 opened Mar 21, 2026 by Thump604 Loading…
4 tasks
5
Use model's generation_config.json for default sampling parameters
#1031 opened Mar 20, 2026 by eyupcanakman Loading…
1
1
Fix per-request enable_thinking toggle in server
#1030 opened Mar 20, 2026 by eyupcanakman Loading…
1
1
Fix CacheDataset.itemlen returning wrong length
#1029 opened Mar 20, 2026 by eyupcanakman Loading…
1
Fix A_log precision in mamba.py
#1028 opened Mar 20, 2026 by eyupcanakman Loading…
1
Stop generation on consecutive duplicate tool calls
#1027 opened Mar 20, 2026 by chopchop-jiahao • Draft
1
Fix SSM dt clamp default for Nemotron-H
#1026 opened Mar 20, 2026 by kernelpool Loading…
1
Feature/slem with context aware
#1025 opened Mar 19, 2026 by krzysiekfonal Loading…
[transformers-to-mlx skill] Add OLMo Hybrid model support
#1023 opened Mar 19, 2026 by pcuenca Loading…
16
ProTip! Updated in the last three days: updated:>2026-03-27.