Skip to content

Pull requests: ml-explore/mlx-lm

New pull request New
110 Open 564 Closed
110 Open 564 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

Add MiMo V2.5
#1219 opened Apr 28, 2026 by kernelpool Contributor Loading…
feat(server): opt-in disk-backed L2 prompt cache (--prompt-cache-disk-dir)
#1218 opened Apr 27, 2026 by freddyhaddad Loading…
Add Metal VJP kernel for gated_delta_update (trainable Qwen3.5 / Qwen3-Next LoRA on Apple Silicon)
#1217 opened Apr 27, 2026 by SudarkinV Loading…
fix(utils): skip already-quantized layers in load_model._quantize predicate
#1216 opened Apr 27, 2026 by adurham Contributor Loading…
Refactor server to use MLXServerConfig
#1213 opened Apr 27, 2026 by giskarda Loading…
1
Add Hy3 preview
#1211 opened Apr 27, 2026 by kernelpool Contributor Loading…
2
Skip quantizing Gemma 4 per_layer_model_projection for Swift compatibility
#1209 opened Apr 27, 2026 by kr1s0404 Loading…
3 tasks
feat(server): add OpenAI Responses API endpoint (/v1/responses)
#1207 opened Apr 27, 2026 by cassiolpaixao90 Loading…
5 tasks
fix(gemma4): drop KV-shared layer projections in sanitize
#1205 opened Apr 26, 2026 by Fox13 Loading…
minimax: validate head_dim against checkpoint, drop unused shared_intermediate_size
#1204 opened Apr 26, 2026 by adurham Contributor Loading…
Add TurboQuantKVCache: 3-bit/4-bit KV cache compression for generation
#1202 opened Apr 26, 2026 by dedalien Loading…
2
Add DeepSeek-V4 (Flash) model support
#1201 opened Apr 26, 2026 by akashgoswami • Draft
3 of 6 tasks
1
Add dense qwen3_5 support for learned quantization
#1200 opened Apr 25, 2026 by iamwavecut Loading…
Add EngGPT MoE model support
#1199 opened Apr 25, 2026 by robertobissanti Loading…
fix: prevent double-shift of norm weights for converted VLM checkpoints
#1198 opened Apr 25, 2026 by Thump604 Loading…
1
feat: add thinking budget with early-stopping prompt injection
#1196 opened Apr 25, 2026 by Thump604 Loading…
2
Implement DSV4
#1195 opened Apr 25, 2026 by rltakashige Contributor Loading…
2
Add DeepSeek-v4 (Flash/Pro)
#1192 opened Apr 24, 2026 by Blaizzy Contributor Loading…
69
feat: add DeepSeek-V4 (Pro/Flash) model support
#1189 opened Apr 24, 2026 by machiabeli Loading…
5 of 7 tasks
30
Include context_length in /v1/models response (#1183)
#1184 opened Apr 23, 2026 by seikixtc Loading…
1
Lc/fix xtc special tokens server
#1176 opened Apr 21, 2026 by micuentadecasa Contributor Loading…
Auto-discover tool-call markers from tokenizer config fields
#1163 opened Apr 18, 2026 by michaelstingl Loading…
6 tasks done
feat(nemotron_h): add Multi-Token Prediction (MTP) module
#1161 opened Apr 16, 2026 by Thump604 Loading…
Add reasoning → tool state machine transition
#1160 opened Apr 16, 2026 by christiangenco Loading…
feature: dynamic quantized model support
#1155 opened Apr 15, 2026 by dsrenesanse • Draft
ProTip! Updated in the last three days: updated:>2026-04-24.