mlx-serve
mlx-serve is 2️⃣下载MLX-Serve App（或brew安装）. It is ranked #91 on the AI Radar, in Open & local models, first seen 52 days ago and shared in 17 posts (203.8k views).
What mlx-serve says about itself
Highlights Qwen 3.8 Flash Next fits on a 64 GB Mac. New 3.3-bit pack ddalcu/Qwen3.8-Flash-Next-MLX-Serve-iQ-MLX-3.3bpw, 52 GB resident, 85.6% top-1 agreement with bf16 (70 GB pack: 89.1%). Its 3-b...
Release v26.9.3 — Flash Next on 64 GB, speculation for everyone, Neural Engine media · ddalcu/mlx-serve · GitHub Skip to content Navigation Menu Sign in Search / Sign in Sign up Appearance settings You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert {{ message }} Repository navigation v26.9.3 — Flash Next on 64 GB, speculation for everyone, Neural Engine media Compare Choose a tag to compare Sorry, something went wrong. Filter Loading Sorry, something…
What people said about mlx-serve on X
🔥牛逼了！12GB显卡本地跑763B！DeepSeek-V4.1-Flash-1017量化版炸裂登场，前沿质量几乎零损失！ ISTA-DASLab团队发布GSQ-RCO非均匀量化版，Strata引擎原生支持。763B参数MoE大模型最低压到1.75 bpw，在消费级硬件上就能跑！ 核心亮点 🔹极致压缩 + 几乎无损：IQ2_XXS（2.00 bpw，约249GB）在AIME25、GPQA-Diamond、LiveCodeBench v6上100%恢复原始模型分数；IQ1_M（1.75 bpw，约231GB）整体恢复98.7%。 🔹真正本地可行：单张12GB显卡 +…
— @NFT_Chen, 13 days ago · 1.7k likes · see the post
DeepSeek V4.1 Flash 1017 Lossless model running on M3 Ultra 512GB. Testing my mlx-serve PR right now @ddalcu 🙌 DeepSeek Harness is top as client! A first round of throughput tests here and a Santa Claus Voxel test. I know David doesn't like these 😂 Note: yes it's faster than DwarfStar, I'll run a comparison soon!…
— @ivanfioravanti, 3 days ago · 129 likes · see the post
My upcoming mlx-serve PR adding support for DeepSeek-V4.1 Flash with tensor parallelism across two Macs over Thunderbolt 5 RDMA. 🔥 Tested on 2× M3 Ultra vs 1× using Lossless weights and llmprobe: • DSpark decode 88 tok/s (vs 56), 116 on predictable text • Plain decode 54 tok/s (vs 41) • Prefill 1,608 tok/s (vs…
— @ivanfioravanti, 2 days ago · 101 likes · see the post
Alternatives to mlx-serve
- kimi-k3.1-pro — Releasing the model weights of Kimi K3.1 Pro
- niklaslenz.com — Play the world in your browser. Every number, the exact command and the raw output
- unsloth.ai — Guide to running Mistral Large 4 GGUFs locally with Unsloth and llama.cpp.
- project-maya — Plus a round of fixes, UI improvements and community PRs.
- twil-8b — Text Generation model with 9B parameters. Download, explore files and run it on Hugging Face.
- deepseek-harness — I have all three of my AI agent leads for Hermes, Grok Bot, and OpenClaw pouring over the repos for DeepSeeks proposed…
mlx-serve in numbers
- Rank on the AI Radar: #91 of 10389
- Shared in 17 posts by 8 accounts: @NFT_Chen, @ivanfioravanti, @ddalcu, @TheDavidTai
- 203.8k views on those posts
- First seen 52 days ago, last shared 1 h ago
- Pricing seen by AGK Intelligence: open source
- Market: Open & local models
FAQ
What is mlx-serve?
2️⃣下载MLX-Serve App（或brew安装） It was first shared on X 52 days ago and is ranked #91 on the AI Radar.
Is mlx-serve free?
It is open source.
Who shared mlx-serve?
8 accounts on X, including @NFT_Chen, @ivanfioravanti, @ddalcu, in 17 posts totalling 203.8k views.
Every post is read and classified by AGK Intelligence: what it is, which market it belongs to, and whether the link is a real tool. 65.7k posts from 5k X accounts over the last 14 days, 10.4k tools, 19 markets. New posts are collected several times an hour and the full ranking is recomputed every hour — last update 2026-11-06 13:30 UTC. Full method.