Nex AGI (Shanghai Innovation Institute), China · Released November 10, 2026New
Nex-N2.6 has 321.3B parameters (17B active per token). In BF16 its weights alone take 585 GiB; quantized to 4-bit, about 176 GiB. Each 32K-token request adds 0.35 GiB of KV cache. Cheapest way to run it today: 4 × RTX A6000 (48 GB) at $1.60 per hour. It reads text, images and video, calls tools and lets you set its reasoning effort.
Read every day from Hugging Face. Split files are added together; vision projectors are left out.
How many GPUs to run Nex-N2.6
Fewest GPUs that fit the weights plus one 8K-token request. Data-center GPUs use vLLM (90% of memory usable), the others llama.cpp. Several GPUs split the model with tensor or pipeline parallelism.
GPU
FP8 (native)
BF16 / FP16
8-bit
4-bit
GeForce RTX 4090 24 GB
Not on this GPU
32 (several machines)
16 (several machines)
16 (several machines)
GeForce RTX 5090 32 GB
Not on this GPU
24 (several machines)
16 (several machines)
6
RTX PRO 6000 Blackwell 96 GB
Not on this GPU
7
4
2
Mac M5 Max 128 GB 96 GB usable
Not on this GPU
7
4
2
Ryzen AI Max+ 395 128 GB 96 GB usable
Not on this GPU
7
4
2
DGX Spark (GB10) 128 GB
Not on this GPU
5
3
2
NVIDIA H100 80 GB
5
16 (several servers)
5
3
NVIDIA H200 141 GB
3
5
3
2
NVIDIA B200 (HGX) 180 GB
2
4
2
2
8-bit: FP8 with vLLM (INT8 on GPUs without FP8), Q8_0 with llama.cpp. 4-bit: AWQ / GPTQ with vLLM, Q4_K_M with llama.cpp, both counted at the Q4_K_M size. Not on this GPU: the official FP8 weights do not run there (llama.cpp needs a GGUF version, and vLLM needs a GPU with FP8 support).
Cheapest way to run Nex-N2.6 today
For each GPU: the fewest cards that fit Nex-N2.6 with one 8K-token request, the most faithful weight format at that count, and the cheapest on-demand price (2026-11-22).
On-demand prices from AWS, Google Cloud, Azure, RunPod, Vast.ai, Oracle Cloud, OVHcloud, Scaleway, Verda, Akash and UpCloud, checked every hour. Serving many users needs more KV cache, so more memory: size it in the calculator. Some links are affiliate links: StudioTV may earn a commission, at no extra cost to you.
vLLM and SGLang apply these defaults unless a request sets its own. With llama.cpp, pass them yourself: --temp 1 --top-p 0.95
Reasoning. Adjustable reasoning effort. Reasoning tokens take context and KV cache like any other token: leave room for them.
KV cache: how memory grows with context
Architecture: 45 layers · MLA (512 + 0 latent) on 11 · indexer keys on 11 · 34 Kimi Delta Attention. Its KV cache is 4.0× smaller than if every layer used full attention, because it keeps only a short window of tokens at full detail and compresses the rest of the sequence. Only 17B parameters are active per token, but all 321B must sit in memory: VRAM depends on the total, speed on the active part.
Context per request
BF16 cache
FP8 cache
8K tokens
0.09 GiB
0.06 GiB
32K tokens
0.35 GiB
0.23 GiB
128K tokens
1.42 GiB
0.93 GiB
1M tokens
11.4 GiB
7.40 GiB
Plus 141 MiB of recurrent state per request, whatever the context.
Nex-N2.6 VRAM FAQ
How much VRAM does Nex-N2.6 need?
In BF16 the weights alone take 585 GiB (628 GB). In FP8 that is 294 GiB, and about 176 GiB with 4-bit quantization (Q4_K_M); the official FP8 checkpoint is 299 GiB. Each request then adds KV cache: 0.09 GiB at 8K tokens and 0.35 GiB at 32K (BF16 cache).
Can Nex-N2.6 run on a single RTX 4090 (24 GB)?
No. Even in 4-bit it needs 16 RTX 4090s, spread over several machines. In 4-bit it needs 3 H100 GPUs.
Can Nex-N2.6 run on an RTX 5090 (32 GB)?
No. Even in 4-bit it needs 6 RTX 5090s.
Can Nex-N2.6 run on a Mac?
Yes. In 4-bit (Q4_K_M with llama.cpp) it needs a Mac with 256 GB of unified memory or more, of which macOS gives the GPU about 192 GB, for one request with an 8K-token context. In 8-bit (Q8_0) it needs 512 GB.
Can Nex-N2.6 run on a DGX Spark or a Ryzen AI Max+ 395 (128 GB)?
No. Even in 4-bit it needs 2 DGX Spark units or 2 Ryzen AI Max+ 395 machines (128 GB), linked over the network.
What is the cheapest way to run Nex-N2.6?
On 2026-11-22, the cheapest on-demand setup is 4 × RTX A6000 (48 GB) with Q4_K_M weights on Vast.ai, at $1.60 per hour (about $1,170.92 per month). Next: 6 × GeForce RTX 5090 (32 GB) with Q4_K_M weights on Vast.ai, at $1.85 per hour. Sized for one 8K-token request; prices are checked every hour.
How many H100 GPUs does Nex-N2.6 need?
With one request and an 8K-token context: 16 in BF16, 5 in FP8 and 3 in 4-bit (vLLM, 90% of memory usable). Above 8 GPUs, that means several servers. Serving many users at once needs more memory for their KV caches: the calculator sizes that for you.
Can Nex-N2.6 read images?
Yes. Nex-N2.6 reads text, images and video. Images add tokens to the context, so they also add KV cache.
Does Nex-N2.6 support tool calling?
Yes, Nex-N2.6 supports tool (function) calling: it can ask your application to run a function and use the result.
Does Nex-N2.6 have a thinking mode?
Adjustable reasoning effort. Reasoning tokens count toward the context and the KV cache.
What temperature should I use with Nex-N2.6?
The official defaults are temperature 1, top-p 0.95.
Nex-N2.6 VRAM badge
For a model card or a README: the 4-bit size and the smallest GPU it fits on, linked to this page. Data as JSON: /api/models/nex-n2-6.json.
Estimates, not guarantees: computed from the official config.json with the same engine as the LLM VRAM Calculator. Real usage depends on your engine version and settings.