Mac Studio M3 Ultra · 96 GB
A quiet single-box option for 70B-class Q4 models and Apple-native MLX workflows, with memory shared by macOS and applications.
Źródło pierwotneQwen3.8 Flash Next REAP-288
The highest listed result may be a tight or offloaded fit. For daily use, prefer the first result marked Swobodnie and keep context modest.
MLX LM
Native Apple silicon inference, experimentation and fine-tuning. Only for Apple silicon. Use an MLX-converted model and tune KV cache size for long contexts. The bundled mlx_lm.server is an OpenAI-compatible HTTP endpoint (default 127.0.0.1:8080), but upstream warns it implements only basic security checks and is not recommended for production.
Otwórz przewodnikReprezentatywne modele open-weight
Qwen3.8 Flash Next REAP-288
- Szacowana pamięć
- ~40.57 GB
- Kontekst
- 7.6K recommended on measured 48 GB Mac
Llama 3.1 70B
- Szacowana pamięć
- ~46 GB
- Kontekst
- 128K
Qwen3.6 35B-A3B
- Szacowana pamięć
- ~25 GB
- Kontekst
- 262K native
Qwen3 32B
- Szacowana pamięć
- ~24.5 GB
- Kontekst
- 32K+
Qwen3 30B-A3B
- Szacowana pamięć
- ~22.5 GB
- Kontekst
- 32K+
Qwen3.6 27B
- Szacowana pamięć
- ~18.5 GB
- Kontekst
- 262K native
Fit is based on estimated total model memory. Driver support, KV cache, multimodal projectors, concurrency, and desktop applications can all reduce available headroom. Current fit: Qwen3.8 Flash Next REAP-288 (Swobodnie), Llama 3.1 70B (Swobodnie), Qwen3.6 35B-A3B (Swobodnie), Qwen3 32B (Swobodnie), Qwen3 30B-A3B (Swobodnie), Qwen3.6 27B (Swobodnie).