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.

nex-agi logo

Nex-N2.6

nex-agi·Instruct·released 10 Nov 2026

Apache 2.0321BA18.3BMoE1M ctx ≈213 GB
Your GPU
ollama run hf.co/bartowski/Nex-N2.6-GGUF:Q4_K_M
All 6 runners ↓

Nex-N2.6 is a multimodal agentic model from Nex-AGI, built on GLM-5.3-Flash. It is designed for computer use, coding, and long-horizon tasks involving tools. This repository contains the FP8-quantized weights of Nex-N2.6, with approximately 321B parameters.

Notify me when

How much VRAM does Nex-N2.6 need?

At Q4_K_M Nex-N2.6 needs about 213 GB with an 8K context (213 GB at Q4, 651 GB at full precision). Pick your GPU to see what fits and how fast it runs.

--vram
any
--gpus
--quant

Needs: ≈219 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 loss117 GB file23 GB≈146 GB
Q3_K_MNoticeable lossSmall, some quality lossNoticeable loss144 GB file23 GB≈175 GB
Q4_K_MrecommendedSmall lossRecommended default for local useSmall loss187 GB file23 GB≈219 GB
Q5_K_MNear-losslessNear-lossless for most tasksNear-lossless227 GB file23 GB≈261 GB
Q6_KNear-losslessPractically losslessNear-lossless262 GB file23 GB≈298 GB
Q8_0LosslessLossless in practice, twice the size of Q4Lossless318 GB file23 GB≈357 GB
FP16 / BF16LosslessFull precision, as releasedLossless599 GB est.23 GB≈651 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 ≈219 GB(187 GB weights + 23 GB KV cache + 9.8 GB overhead). Pick a GPU above to see how much context fits.

010002000300040001K2K4K8K16K32K64K128K256K512K1MGB

Weights + overheadKV cache

Can the Apple M3 Ultra (512 GB) run Nex-N2.6? →

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 Nex-N2.6

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 Nex-N2.6 locally

--quantQ4_K_Mdefault; pick your GPU to tailor itbartowski/Nex-N2.6-GGUF

Q4_K_M straight from Hugging Facedocs
ollama run hf.co/bartowski/Nex-N2.6-GGUF:Q4_K_M

Install Ollama

OpenAI-compatible server with a web UIdocs
llama-server -hf bartowski/Nex-N2.6-GGUF:Q4_K_M

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 bartowski/Nex-N2.6-GGUF@Q4_K_M

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

Install LM Studio

Serve an OpenAI-compatible APIdocs
vllm serve nex-agi/Nex-N2.6
Docker Compose (NVIDIA)docs
yaml
services:
  vllm:
    image: vllm/vllm-openai:latest
    command: nex-agi/Nex-N2.6
    ports:
      - "8000:8000"
    ipc: host
    volumes:
      - ~/.cache/huggingface:/root/.cache/huggingface
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

API on port 8000.

Install vLLM

Serve an OpenAI-compatible APIdocs
sglang serve nex-agi/Nex-N2.6 --host 0.0.0.0 --port 30000
Docker Compose (NVIDIA)docs
yaml
services:
  sglang:
    image: lmsysorg/sglang:latest
    command: sglang serve nex-agi/Nex-N2.6 --host 0.0.0.0 --port 30000
    ports:
      - "30000:30000"
    ipc: host
    volumes:
      - ~/.cache/huggingface:/root/.cache/huggingface
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]

API on port 30000.

Install SGLang

Ask about an imagedocs
python
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="nex-agi/Nex-N2.6", 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
bartowski/Nex-N2.6-GGUFGGUFIQ1_S75 GB8.1K
bartowski/Nex-N2.6-GGUFGGUFIQ1_M83 GB8.1K
bartowski/Nex-N2.6-GGUFGGUFIQ2_XXS97 GB8.1K
bartowski/Nex-N2.6-GGUFGGUFIQ2_XS101 GB8.1K
bartowski/Nex-N2.6-GGUFGGUFIQ2_S108 GB8.1K
bartowski/Nex-N2.6-GGUFGGUFIQ2_M121 GB8.1K
bartowski/Nex-N2.6-GGUFGGUFQ2_K126 GB8.1K
bartowski/Nex-N2.6-GGUFGGUFIQ3_XXS139 GB8.1K
bartowski/Nex-N2.6-GGUFGGUFIQ3_XS147 GB8.1K
bartowski/Nex-N2.6-GGUFGGUFQ3_K_S147 GB8.1K

About Nex-N2.6

Nex-N2.6 is a multimodal agentic model from Nex-AGI, built on GLM-5.3-Flash. It is designed for computer use, coding, and long-horizon tasks involving tools. This repository contains the FP8-quantized weights of Nex-N2.6, with approximately 321B parameters. Following Nex-N2.5-Pro, this release focuses on reducing repetitive reasoning and improving adherence to the tools available in the current environment. • 🔥 Less repetitive reasoning: Reasoning repetition fell from 16.0% to 1.6% compared with Nex-N2.5-Pro in our internal checkpoint-replay evaluation—a 90% relative reduction. • 🛠️ Improved tool-list adherence: Nex-N2.6 achieved 100% tool-list adherence among responses making tool calls in our internal evaluation. • 🖥️ Stronger computer use: Nex-N2.6 scored 83.0 on OSWorld, up from 82.2 for Nex-N2.5-Pro and 78.0 for the base model, GLM-5.3-Flash. The following results are reported by the Nex-AGI team for this release. Scores depend on the evaluation harness and settings. These are model-level results, not a separate evaluation of this FP8 export.

Full model card on Hugging Face →

Alternatives to Nex-N2.6

All alternatives →
zai-org logo

GLM-5.3-Flash

zai-org·Vision

≈213 GB

👋 Join our WeChat or Discord community. 📖 Check out the GLM-5.3-Flash blog and GLM-5 Technical report . 📍 Use GLM-5.3-Flash API services on Z.ai API Platform.

321BA18.3B1M ctx MIT
6.4MAug 2026ollama
thinkingmachines logo

Inkling-Small

thinkingmachines·Vision

≈159 GB

Inkling-Small 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.

266BA1.7B1M ctx Apache 2.0
916.1KJul 2026gguf
zai-org logo

GLM-5.3-Flash-BF16

zai-org·Vision

≈213 GB

👋 Join our WeChat or Discord community. 📖 Check out the GLM-5.3-Flash blog and GLM-5 Technical report . 📍 Use GLM-5.3-Flash API services on Z.ai API Platform.

321BA18.3B1M ctx MIT
21.9KAug 2026gguf
deepseek-ai logo

DeepSeek-V4-Flash-Vision-Exp

deepseek-ai·Vision

≈182 GB

We are excited to introduce DeepSeek-V4-Flash-Vision-Exp, our first experimental multimodal model in the DeepSeek-V4 family. It builds on the DeepSeek-V4-Flash architecture by incorporating visual modules and undergoing continued training to unlock visual understanding capabilities.

305BA15.2B1M ctx MIT
30.6KAug 2026gguf
Qwen logo

Qwen3.5-397B-A17B

Qwen·Vision

≈241 GB

Over recent months, we have intensified our focus on developing foundation models that deliver exceptional utility and performance.

403BA17B256K ctx Apache 2.0
645.4KFeb 2026gguf
Qwen logo

Qwen3.8-Flash-Next

Qwen·Vision

≈108 GB

As the frontier of foundation models pushes toward ever-larger parameter counts and ever-longer context windows, the question is no longer just how much we can scale, but how efficiently we can do so.

180BA61.6B256K ctx qwen-community-1.0
2.7MAug 2026ollama

Nex-N2.6 FAQ

How much VRAM does Nex-N2.6 need?
About 213 GB at Q4_K_M with an 8K context. Q8_0 needs about 357 GB and full precision about 651 GB. Longer contexts need more memory for the KV cache.
Can Nex-N2.6 run on the Apple M3 Ultra (512 GB)?
Yes. It fits at Q8_0 (about 357 GB), at an estimated 23 tokens/s.
Can I use Nex-N2.6 commercially?
Yes. It is released under Apache 2.0. Permissive: use, modify and sell freely; keep the notice and licence text.
What is the context length of Nex-N2.6?
1,048,576 tokens (1M). The full window needs far more memory than an 8K context.
How do I run Nex-N2.6 with Ollama or llama.cpp?
With Ollama: ollama run hf.co/bartowski/Nex-N2.6-GGUF:Q4_K_M. With llama.cpp: llama-server -hf bartowski/Nex-N2.6-GGUF:Q4_K_M, 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.