Yes — DeepSeek V4 Flash Vision-Exp (305B-A13B) runs at 33 tok/s on M3 Ultra with 256 GB RAM using Q4_K_M quantization via llama.cpp. A capable open-source LLM with 305B-A13B parameters.
Want to run DeepSeek V4 Flash Vision-Exp faster — or step up to bigger models? Find your Mac →The LLMCheck index estimates DeepSeek V4 Flash Vision-Exp on M3 Ultra using our published methodology: Q4_K_M quantization, memory-bandwidth scaling, and cross-referenced third-party benchmarks where available. Figures are transparent estimates — own this config? Submit a real benchmark →
| Metric | Value |
|---|---|
| Tokens per second | 33 tok/s |
| Time to first token | — |
| Quantization | Q4_K_M |
| Minimum RAM | 256 GB |
| Recommended engine | llama.cpp |
| Parameters | 305B-A13B |
| Benchmark date | 2026-09 |
Q4_K_M 305B-A13B llama.cpp M3 Ultra
The recommended engine for DeepSeek V4 Flash Vision-Exp on M3 Ultra is llama.cpp. Install Ollama, then pull the model:
Ollama handles quantization automatically — it will download the Q4_K_M variant (~256 GB) and start an interactive chat session.
| Chip | Speed | First Token | Min RAM | Engine |
|---|---|---|---|---|
| M5 Ultra | 47 tok/s | — | 256 GB | llama.cpp |
| M4 Ultra | 42 tok/s | — | 192 GB | llama.cpp |
To run DeepSeek V4 Flash Vision-Exp on M3 Ultra you need:
DeepSeek V4 Flash Vision-Exp needs about 256 GB of unified memory. These current Apple Silicon Macs have the headroom to run it comfortably:
As an Amazon Associate, LLMCheck earns from qualifying purchases. These affiliate links cost you nothing extra and help keep our benchmarks free.
See how DeepSeek V4 Flash Vision-Exp stacks up against other models on your specific Mac hardware.
Open Compare Tool Full Leaderboard