Ministral 3 14B
Ministral 3 14B needs 10.3 GB of VRAM at Q4_K_M with 8k context, computed from its config.json as checked on 3 November 2026. The smallest card that holds it has 12 GB, a tight fit, and the fastest desktop card that runs it, the RTX 5090, generates at most 154 tokens per second.
- Parameters
- 14B
- Context
- 262,144 tokens
- Licence
- Apache 2.0
- Released
- December 2025
- Config checked
- 3 November 2026
Open Ministral 3 14B in the LLM checker
Which card size runs it
10.3 GB against each memory size. A model may use 95% of a card, and only what is left over, or missing, is coloured; tight means above 85%, with little room for a longer conversation.
Other quantisations, context lengths or a Q8 KV cache: the VRAM calculator.
Computed 3 November 2026Which GPU to pick
The smallest card here that loads it at Q4_K_M with 8k of context is the RTX 3060 (12 GB): 10.3 GB needed, an estimated 31 tokens per second, and a tight fit that leaves little for the conversation. For room to work, meaning 32,768 tokens of context while staying under 85% of memory, step up to the RX 7900 XT (20 GB), which needs 14.3 GB for that and generates at about 69 tokens per second.
At Q8_0, which is near-lossless, the card is the RX 7900 XT (20 GB), needing 17.3 GB. On Apple silicon, a Mac mini M4 Pro holds it with 32,768 tokens of context at an estimated 24 tokens per second; prompt processing is slower than on Nvidia, which long documents make obvious. Among unified-memory desktops, the Ryzen AI Max+ 395 holds it at 32,768 tokens in 14.3 GB, at an estimated 22 tokens per second.
Speed on each card
| GPU | Fits | Est. tokens/s | Feels like | Memory | Bandwidth |
|---|---|---|---|---|---|
| H100 SXM | Yes | ≤ 289 | faster than you can read | 80 GB | 3,350 GB/s |
| A100 | Yes | ≤ 176 | faster than you can read | 80 GB | 2,039 GB/s |
| RTX 5090 | Yes | ≤ 154 | faster than you can read | 32 GB | 1,792 GB/s |
| RTX PRO 6000 Blackwell | Yes | ≤ 154 | faster than you can read | 96 GB | 1,792 GB/s |
| RTX PRO 5000 Blackwell | Yes | ≤ 116 | faster than you can read | 48 GB | 1,344 GB/s |
| RTX 4090 | Yes | 87 | faster than you can read | 24 GB | 1,008 GB/s |
| RTX 5080 | Yes | 83 | faster than you can read | 16 GB | 960 GB/s |
| RX 7900 XTX | Yes | 83 | faster than you can read | 24 GB | 960 GB/s |
| RTX 6000 Ada | Yes | 83 | faster than you can read | 48 GB | 960 GB/s |
| RTX 3090 | Yes | 81 | faster than you can read | 24 GB | 936 GB/s |
| RTX 5070 Ti | Yes | 77 | faster than you can read | 16 GB | 896 GB/s |
| RTX 5090 Laptop GPU | Yes | 77 | faster than you can read | 24 GB | 896 GB/s |
| RTX PRO 4500 Blackwell | Yes | 77 | faster than you can read | 32 GB | 896 GB/s |
| L40S | Yes | 74 | faster than you can read | 48 GB | 864 GB/s |
| RX 7900 XT | Yes | 69 | faster than you can read | 20 GB | 800 GB/s |
| RTX A6000 | Yes | 66 | faster than you can read | 48 GB | 768 GB/s |
| RTX 4080 SUPER | Yes | 63 | faster than you can read | 16 GB | 736 GB/s |
| RTX 5070 | Tight | 58 | comfortable for chat | 12 GB | 672 GB/s |
| RTX PRO 4000 Blackwell | Yes | 58 | comfortable for chat | 24 GB | 672 GB/s |
| RX 9070 XT | Yes | 55 | comfortable for chat | 16 GB | 640 GB/s |
| Radeon AI PRO R9700 | Yes | 55 | comfortable for chat | 32 GB | 640 GB/s |
| Arc Pro B70 | Yes | 52 | comfortable for chat | 32 GB | 608 GB/s |
| RTX 4070 | Tight | 43 | comfortable for chat | 12 GB | 504 GB/s |
| Arc B580 | Tight | 39 | comfortable for chat | 12 GB | 456 GB/s |
| Arc Pro B60 | Yes | 39 | comfortable for chat | 24 GB | 456 GB/s |
| RTX 5060 Ti | Yes | 39 | comfortable for chat | 16 GB | 448 GB/s |
| RTX PRO 4000 Blackwell SFF | Yes | 37 | comfortable for chat | 24 GB | 432 GB/s |
| RTX 3060 | Tight | 31 | comfortable for chat | 12 GB | 360 GB/s |
| RX 9060 XT | Yes | 28 | comfortable for chat | 16 GB | 320 GB/s |
| RTX Spark | Yes | 26 | comfortable for chat | 128 GB (102 for the GPU) | 300 GB/s |
| RTX 4060 Ti | Yes | 25 | usable, a little slow | 16 GB | 288 GB/s |
| DGX Spark | Yes | 24 | usable, a little slow | 128 GB (126 for the GPU) | 273 GB/s |
| Ryzen AI Max+ 395 | Yes | 22 | usable, a little slow | 128 GB (96 for the GPU) | 256 GB/s |
Single-stream decode, from memory bandwidth at 70% efficiency. Prompt processing and batching not included. These are ceilings, not measurements: Ministral 3 14B reads only 8.1 GB of weights per token at Q4, so little that the per-token costs the formula leaves out decide much of the real speed. The 5 figures marked ≤ are where a real runtime falls furthest below the number shown. See the speed estimator for other quantisations and Apple hardware, or every card compared if you are choosing hardware rather than a model.
Computed 3 November 2026Memory by quantisation and context
| Quant | 4,096 ctx | 8,192 ctx | 32,768 ctx |
|---|---|---|---|
| Q8_0 | 16.6 GB · fits 24 GB | 17.3 GB · fits 24 GB | 21.3 GB · fits 24 GB |
| Q5_K_M | 11.5 GB · fits 16 GB | 12.2 GB · fits 16 GB | 16.2 GB · fits 24 GB |
| Q4_K_M | 9.6 GB · fits 12 GB | 10.3 GB · fits 12 GB | 14.3 GB · fits 16 GB |
Weights at this quant: 8.1 GB at Q4. Every extra 1,000 tokens of context adds about 0.164 GB of KV cache at FP16. Try other settings in the VRAM calculator.
About the model
Ministral 3 14B has 14B parameters, 40 layers and 8 KV heads of dimension 128. The largest of Mistral's Ministral 3 family, December 2025, designed for edge and local deployment. A conventional dense transformer with vision, a 256k window and an Apache licence.
Text and images 262,144-token window Apache 2.0 December 2025
Reasoning. The Instruct release answers directly. Mistral publishes a separate Reasoning variant of the same size.
The parameter count includes the vision encoder; a text-only GGUF is slightly smaller.
Native context window: 262,144 tokens. Licence: Apache 2.0, a permissive licence. Weights published in December 2025.
Good at
- A 16 GB card: it fits with a useful context, and Mistral ships it in FP8 for cards that support it.
- Function calling and JSON output, which the card lists as design goals.
- European-language work: French, Spanish, German, Italian, Portuguese and Dutch are named explicitly.
Watch out for
- Every layer keeps a full-length cache, so long context costs several times what a 2026 hybrid model pays for the same conversation.
- The 256k window is real, but filling it costs more memory than the weights do.
Similar models
The models that need about the same memory as Ministral 3 14B at Q4_K_M and 8k, which makes them the real alternatives on whatever card you have. The context column is what each extra 1,000 tokens costs, and it is where models of the same size differ most.
| Model | Size | Needs | Per 1k context | Window | Released |
|---|---|---|---|---|---|
| Ministral 3 14B | 14B | 10.3 GB | 0.164 GB | 256k | 2025-12 |
| Qwen3 14B | 14.8B | 10.8 GB | 0.164 GB | 40k | 2025-04 |
| Phi-4 14B | 14.7B | 11 GB | 0.205 GB | 16k | 2024-12 |
| Gemma 3 12B | 12.2B | 8.7 GB | 0.066 GB | 128k | 2025-03 |
| Gemma 4 12B | 12B | 8.2 GB | 0.016 GB | 256k | 2026-05 |
See every model compared, which models fit on 8 to 96 GB, or the cheapest hour of every cloud card.
Questions
- Can I run Ministral 3 14B on a 24 GB card?
- Yes, at Q4_K_M and 8k context it needs about 10.3 GB, which fits a 24 GB card with room.
- How much VRAM does Ministral 3 14B need at Q8?
- About 17.3 GB at 8k context, or 21.3 GB at 32k. Q8 is near-lossless; use it when it fits.
- What is the context length of Ministral 3 14B?
- 262,144 tokens natively. Holding all of it at Q4_K_M takes about 51.9 GB, of which 42.9 GB is context.
- How fast is Ministral 3 14B on an RTX 4090?
- At most about 87 tokens per second at Q4, single stream, which is faster than you can read. That is a ceiling worked out from the card's memory bandwidth, not a measurement.
- Architecture values from config.json in mistralai/Ministral-3-14B-Instruct-2512 on Hugging Face, checked 3 November 2026. Every figure on this page is computed from those values; check them before buying hardware for this model.
- Estimates assume a single conversation on a card that is otherwise free. A desktop on the same GPU takes 0.5 to 2 GB.