selfhost.directory

The index of self-hostable projects — ranked, categorized and served with live updates.
Own your data and stop giving it to big tech.

mistralai logo

Mistral-Large-4-1T-A52B

mistralai·Instruct·released 30 Sept 2026

Custom licence1TA52BMoE1M ctx ≈602 GB
Your GPU
ollama run hf.co/unsloth/Mistral-Large-4-1T-A52B-GGUF:UD-Q4_K_XL
All 4 runners ↓

Mistral Large 4 is our largest and most capable model to date: a 1 trillion-parameter natively multimodal Mixture-of-Experts model with 49 billion active parameters per token (52 billion including embeddings and output layers), trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's…

Notify me when

How much VRAM does Mistral-Large-4-1T-A52B need?

At Q4_K_M Mistral-Large-4-1T-A52B needs about 602 GB with an 8K context (602 GB at Q4, 1965 GB at full precision). Pick your GPU to see what fits and how fast it runs.

--vram
any
--gpus
--quant

Needs: ≈602 GB · Q4_K_M · 8K context — Pick a GPU to check the fit.

Change quant, context or KV cache to compare
Advanced: context 8K · KV cache 16-bit
Context
KV cache

Smallest GPU for each quant and context (KV cache 16-bit)

Quantisation8K16K32K64K128K1M
Q3_K_M
Q4_K_M
Q5_K_M
Q6_K
Q8_0
FP16

≤16 GB card24–32 GB48–96 GBmore

QuantisationQualityWeightsKV cacheTotal
Q2_KHeavy lossSmallest, noticeably worse qualityHeavy loss306 GB est.9.0 GB≈331 GB
Q3_K_MNoticeable lossSmall, some quality lossNoticeable loss455 GB est.9.0 GB≈487 GB
Q4_K_MrecommendedSmall lossRecommended default for local useSmall loss565 GB est.9.0 GB≈602 GB
Q5_K_MNear-losslessNear-lossless for most tasksNear-lossless662 GB est.9.0 GB≈705 GB
Q6_KNear-losslessPractically losslessNear-lossless767 GB est.9.0 GB≈815 GB
Q8_0LosslessLossless in practice, twice the size of Q4Lossless1039 GB file9.0 GB≈1100 GB
FP16 / BF16LosslessFull precision, as releasedLossless1955 GB file9.0 GB≈2062 GB

Estimates for 8K tokens of context: weights + KV cache + runtime overhead. All experts stay in memory, but only the active ones are read per token. Speed assumes memory-bound decoding at your GPU’s bandwidth. Pick a GPU above and the fit column checks every row against it. Weights marked “file” use the real GGUF download size; “est.” rows are computed from the parameter count.

How to pick a quant

Take the biggest quant that fits with room left for context. Q4_K_M is the usual sweet spot. Below Q4 quality drops fast, above Q6 you mostly pay memory for nothing. Big models (70B+, large MoE) tolerate low quants better than small ones.

  • LosslessF16, Q8_0
  • Near-losslessQ6_K, Q5_K_M
  • Small lossQ5_K_S, Q4_K_M, IQ4_NL, IQ4_XS, Q4_K_S
  • Noticeable lossQ4_0, Q3_K_L, Q3_K_M, IQ3_M
  • Heavy lossIQ3_XXS, Q3_K_S, IQ2_M, Q2_K

Typical quality loss, not measured per model. Tiers come from KL divergence against FP16 on Llama 3 8B: llama.cpp perplexity README: LLaMA 3 8B scoreboard (KLD and ΔPPL vs FP16, Wikitext-2); llama.cpp quantize tool: ΔPPL per quant type @ Llama-3-8B

How much context fits?

Model quant
KV cache

At 8K context, Q4_K_M needs ≈602 GB(565 GB weights + 9.0 GB KV cache + 29 GB overhead). Pick a GPU above to see how much context fits.

05001000150020001K2K4K8K16K32K64K128K256K512K1MGB

Weights + overheadKV cache

Speed vs quality per quant

Smaller quants run faster and fit in less memory, but lose quality.

Q2_KQ3_K_MQ4_K_MQ5_K_MQ6_KQ8_0F16

Fits in VRAMTight fitCPU offload✕ won't runQuality is typical KL divergence vs FP16, not measured for this model.

Real-world speed of Mistral-Large-4-1T-A52B

Tokens per second measured by people running it at home. A GPU shows a median once 3 reports agree on the same setup. Reports far from our estimate wait for a review. Measure yours with llama-bench →

No medians yet. Be one of the first 3 to report a setup.

All reports

Run Mistral-Large-4-1T-A52B locally

--quantQ4_K_Mdefault; pick your GPU to tailor itunsloth/Mistral-Large-4-1T-A52B-GGUF

UD-Q4_K_XL straight from Hugging Facedocs
ollama run hf.co/unsloth/Mistral-Large-4-1T-A52B-GGUF:UD-Q4_K_XL

Install Ollama

OpenAI-compatible server with a web UIdocs
llama-server -hf unsloth/Mistral-Large-4-1T-A52B-GGUF:UD-Q4_K_XL

Downloads the GGUF on first run. Installed through llama.app? The same flags work with `llama serve`.

Install llama.cpp

Download with the lms CLIdocs
lms get unsloth/Mistral-Large-4-1T-A52B-GGUF@UD-Q4_K_XL

Then load it from the LM Studio app or with `lms load`.

Install LM Studio

Ask about an imagedocs
python
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="mistralai/Mistral-Large-4-1T-A52B", dtype="auto", device_map="auto")
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/bee.jpg"},
            {"type": "text", "text": "Describe this image."},
        ],
    }
]
out = pipe(text=messages, max_new_tokens=256)
print(out[0]["generated_text"][-1]["content"])

Install Transformers

Quantised versions and downloads

RepositoryFormatQuantSizeDownloads
unsloth/Mistral-Large-4-1T-A52B-GGUFGGUFUD-IQ1_S233 GB20.8K
unsloth/Mistral-Large-4-1T-A52B-GGUFGGUFUD-IQ1_M260 GB20.8K
unsloth/Mistral-Large-4-1T-A52B-GGUFGGUFUD-IQ2_XXS302 GB20.8K
unsloth/Mistral-Large-4-1T-A52B-GGUFGGUFUD-Q2_K_XL378 GB20.8K
unsloth/Mistral-Large-4-1T-A52B-GGUFGGUFUD-IQ3_XXS424 GB20.8K
unsloth/Mistral-Large-4-1T-A52B-GGUFGGUFUD-Q3_K_XL493 GB20.8K
unsloth/Mistral-Large-4-1T-A52B-GGUFGGUFUD-Q4_K_XL605 GB20.8K
unsloth/Mistral-Large-4-1T-A52B-GGUFGGUFUD-Q5_K_XL784 GB20.8K
unsloth/Mistral-Large-4-1T-A52B-GGUFGGUFUD-Q6_K_XL890 GB20.8K
unsloth/Mistral-Large-4-1T-A52B-GGUFGGUFQ8_01115 GB20.8K

About Mistral-Large-4-1T-A52B

Mistral Large 4 is our largest and most capable model to date: a 1 trillion-parameter natively multimodal Mixture-of-Experts model with 49 billion active parameters per token (52 billion including embeddings and output layers), trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's own datacenters in Europe. Mistral Large 4 unifies instruction following, reasoning, coding, and agentic workflows in a single set of weights. While a generalist model, it shows especially good performance on workloads such as cybersecurity, finance, and law, and it surpasses frontier closed models on visual grounding. It is natively fluent in 160+ languages, including every official language of the European Union. Find more information on our blog. The model is also served on the Mistral API as mistral-large-4. Mistral Large 4 includes the following architectural choices: • ~1T total / ~52B active parameters: granular Mixture-of-Experts with 128 routed experts (4 active per token) and 1 shared expert. • 1M context length, for long-context workloads. • Multimodal input: accepts text and image input, with text output. Pixtral-style vision encoder trained to handle variable image sizes and aspect ratios. • Instruct and Reasoning functionalities with function calls, reasoning effort configurable per request. Mistral Large 4 offers the following capabilities:

Full model card on Hugging Face →

Alternatives to Mistral-Large-4-1T-A52B

All alternatives →
thinkingmachines logo

Inkling

thinkingmachines·Vision

≈567 GB

Inkling is a general-purpose multimodal model that accepts text, image and audio inputs and generates text outputs. It is intended for use in English and other languages, and across multiple coding languages.

952BA18.1B1M ctx Apache 2.0
1.4MJul 2026gguf
deepseek-ai logo

DeepSeek-V4.1-Flash

deepseek-ai·Vision

≈454 GB

We introduce DeepSeek-V4.1-Flash, a multimodal Mixture-of-Experts (MoE) model with 552B backbone parameters and support for contexts of up to one million tokens. The model natively processes images and text, and generates text autoregressively.

763BA188B1M ctx MIT
1.5MSept 2026gguf
moonshotai logo

Kimi-K2.7-Code

moonshotai·Vision

≈632 GB

Kimi K2.7 Code is a coding-focused agentic model built upon Kimi K2.6. With substantial improvements on real-world long-horizon coding tasks, it strengthens end-to-end task completion across complex software engineering workflows while improving token efficiency, reducing thinking-token usage by…

1.03TA33.3B256K ctx modified-mit
321.9KJun 2026ollama
deepseek-ai logo

DeepSeek-V4.1-Flash-1017

deepseek-ai·Vision

≈454 GB

DeepSeek-V4.1-Flash-1017 is a minor update to DeepSeek-V4.1-Flash, delivering substantial improvements across reasoning, coding, agentic tasks, and visual benchmarks.

763BA188B1M ctx MIT
493KOct 2026gguf
moonshotai logo

Kimi-K2.6

moonshotai·Vision

≈632 GB

Kimi K2.6 is an open-source, native multimodal agentic model that advances practical capabilities in long-horizon coding, coding-driven design, proactive autonomous execution, and swarm-based task orchestration.

1.03TA33.3B256K ctx modified-mit
202.3KApr 2026ollama
moonshotai logo

Kimi-K2.5

moonshotai·Vision

≈632 GB

Kimi K2.5 is an open-source, native multimodal agentic model built through continual pretraining on approximately 15 trillion mixed visual and text tokens atop Kimi-K2-Base.

1.03TA33.3B256K ctx modified-mit
394.7KJan 2026gguf

Mistral-Large-4-1T-A52B FAQ

How much VRAM does Mistral-Large-4-1T-A52B need?
About 602 GB at Q4_K_M with an 8K context. Q8_0 needs about 1049 GB and full precision about 1965 GB. Longer contexts need more memory for the KV cache.
Can Mistral-Large-4-1T-A52B run on a consumer GPU?
Not comfortably on a single one: it needs about 602 GB at Q4. Split it over several GPUs, use a large unified-memory machine, or offload part of it to system RAM.
Can I use Mistral-Large-4-1T-A52B commercially?
Unclear. It is released under Custom licence. Custom licence: read the model card before using it commercially. Check the licence on the model card before using it in a product.
What is the context length of Mistral-Large-4-1T-A52B?
1,048,576 tokens (1M). The full window needs far more memory than an 8K context.
How do I run Mistral-Large-4-1T-A52B with Ollama or llama.cpp?
With Ollama: ollama run hf.co/unsloth/Mistral-Large-4-1T-A52B-GGUF:UD-Q4_K_XL. With llama.cpp: llama-server -hf unsloth/Mistral-Large-4-1T-A52B-GGUF:UD-Q4_K_XL, which starts an OpenAI-compatible server.

Data from Hugging Face, the model card and public leaderboards. Memory and speed figures are estimates; real numbers depend on the runtime, context length and drivers. Always check the licence yourself before commercial use.