Weekly archive

New local AI models, Oct 13–19, 2025

2 open models released Oct 13–19, 2025, with how each runs on five reference devices.

Qwen3-VL 32B Instruct

Alibaba · LLM · 33B · released Oct 19

74Great quality (A) · 74 of 100: how capable the model is at Q4_K_M, whatever it runs on. Match score 73/100 (quality, fit, speed and memory use) sets the ranking.

Runs on
- Phone iPhone 17 Pro Won't runover the app memory limit
- Laptop MacBook Pro M5 Pro 64 GB Runs well 12.6 tok/s
- Gaming PC RTX 5090 32 GB Runs welltight fit 45.8 tok/s
- Edge board Jetson Orin NX 16 GB Won't run
- Workstation Mac Studio 512 GB Runs well 26.0 tok/s
Memory, every quant and how to run it Check your own device
Qwen3-VL 2B Instruct

Alibaba · LLM · 2.1B · released Oct 19

36Basic quality (D) · 36 of 100: how capable the model is at F16, whatever it runs on. Match score 58/100 (quality, fit, speed and memory use) sets the ranking.

Runs on
- Phone iPhone 17 Pro Runs great 45.5 tok/s
- Laptop MacBook Pro M5 Pro 64 GB Runs great 77.6 tok/s
- Gaming PC RTX 5090 32 GB Runs great 242 tok/s
- Edge board Jetson Orin NX 16 GB Runs great 40.9 tok/s
- Workstation Mac Studio 512 GB Runs great 95.4 tok/s
Memory, every quant and how to run it Check your own device

Estimates come from published hardware specs and our open formulas, not from running the model on your machine. Real speed varies with runtime, drivers and settings. Numbers marked “Measured” come from a linked benchmark.

Qwen3-VL 32B Instruct

Qwen3-VL 2B Instruct