Ministral 3 14B
Ministral 3 14B is an open-weights language model from Mistral, released December 2, 2025. It has 14B parameters (dense). The context window is 256K tokens. API price: $0.2 input / $0.2 output per million tokens.
Today, we announce Mistral 3, the next generation of Mistral models. Mistral 3 includes three state-of-the-art small, dense models (14B, 8B, and 3B) and Mistral Large 3 – our most capable model to date – a sparse mixture-of-experts trained with 41B active and 675B total parameters.
Check it against your own PC →
Can you run Ministral 3 14B locally?
The smallest common GPU it runs on is 6 GB of VRAM (with 32 GB of RAM): Q3_K_M, split GPU / CPU, roughly 5 tokens per second. Estimates assume 32 GB of DDR5 RAM, a 16K context and llama.cpp-style GGUF quantization. Enter your exact hardware in the live app.
| GPU memory | Verdict | Best quant | VRAM | RAM | tok/s |
|---|---|---|---|---|---|
| 6 GB VRAM | Runs, slowly | Q3_K_M (okay) | 5.2 GB | 4.9 GB | ~5 |
| 8 GB VRAM | Runs, slowly | Q4_K_M (very good) | 7.2 GB | 4.6 GB | ~6 |
| 12 GB VRAM | Runs on GPU | IQ4_XS (good) | 10.7 GB | – | ~28 |
| 16 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 14.7 GB | – | ~26 |
| 24 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 14.7 GB | – | ~47 |
| 32 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 14.7 GB | – | ~86 |
Ready-made quantizations on Hugging Face: mradermacher/Ministral-3-14B-Base-2512-i1-GGUF · Mungert/Ministral-3-14B-Reasoning-2512-GGUF · mradermacher/Ministral-3-14B-Base-2512-GGUF · cafepm/Ministral-3-14B-Instruct-2512-Q3Q4-GGUF · mlx-community/Ministral-3-14B-Instruct-2512-4bit · mlx-community/Ministral-3-14B-Reasoning-2512-4bit
Benchmark results
Epoch AI has not published results for Ministral 3 14B yet. New models are usually evaluated within days to weeks.
Links
Announcement · Hugging Face: mistralai/Ministral-3-14B-Base-2512
More from Mistral
Mistral Large 4 · Mistral Large 4 Preview · Mistral Large 4 1T A52B · Leanstral 1.5 · Mistral Medium 3.5 · Mistral Small 4 · Devstral Small 2 · Devstral 2