33B Nanbeige
Nanbeige4.3 33B is a dense 33B model in the plain Llama layout, released under Apache 2.0. Its 512K window comes from a very large RoPE base, and every one of its 52 layers keeps a full KV cache (8 KV heads), so context is expensive: about 213 KB per token. Q4_K_M needs about 22.6 GB at 4K, tight on a 24 GB card, and 25.3 GB at 16K, which no longer fits one. Q3_K_M needs about 17.5 GB at 4K, comfortable on a 24 GB card, and 20.2 GB at 16K. A 32 GB card holds Q4_K_M comfortably to 16K. The full 512K window needs well over 100 GB of cache alone. Sizes come from bartowski's GGUF files; no per-level quality loss has been published.
524K
Max Context
3
Quant Variants
GGUF Q8_0
Best Quality
—
Perplexity rise (best build)
Quantization Variants #
Per-quant VRAM, quality loss, and inference speed on RTX 4090
Measured = run on this site · Estimated = not run here: calculated, or a figure this site has not verified · Community = public reports
| Format | Level | BPW | VRAM | PPL Loss | Speed | Source | Actions |
|---|---|---|---|---|---|---|---|
| GGUF | Q3_K_M | 3.85 | 17.6 GB | — | — | Estimated | |
| GGUF | Q4_K_M | 5.02 | 22.6 GB | — | — | Estimated | |
| GGUF | Q8_0 | 8.5 | 37.7 GB | — | — | Estimated |
Which file each size is for
A file size is not a memory figure: the VRAM column is the calculator's total at 4K context — weights plus KV cache plus a 10% activation buffer. Units differ on purpose: file sizes are decimal GB, as Hugging Face lists them; the VRAM column is GiB (1024³ bytes) with 2% added to the weights, so 7.12 GB of file counts as about 6.8 GiB there. A vision projector (mmproj), where the repo ships one, is a separate download and is in neither number.
- GGUF Q3_K_Mbartowski/Nanbeige_Nanbeige4.3-33B-GGUF · Nanbeige_Nanbeige4.3-33B-Q3_K_M.gguf · 15.92 GB (15,919,746,176 bytes) · listing read 2026-10-26
- GGUF Q4_K_Mbartowski/Nanbeige_Nanbeige4.3-33B-GGUF · Nanbeige_Nanbeige4.3-33B-Q4_K_M.gguf · 20.79 GB (20,788,825,216 bytes) · listing read 2026-10-26
- GGUF Q8_0bartowski/Nanbeige_Nanbeige4.3-33B-GGUF · Nanbeige_Nanbeige4.3-33B-Q8_0.gguf · 35.20 GB (35,195,128,960 bytes) · listing read 2026-10-26
Similar models #
[Compare with Qwen3-VL 30B-A3B](https://quantized.uk/tools/compare/?a=nanbeige4.3-33b&b=qwen3-vl-30b-a3b)
[30B-A3BConsumer GPUPro GPU15.8 GBmin VRAM+0.3%perplexity rise](https://quantized.uk/quant-hub/qwen3-vl-30b-a3b/)
### Qwen3-VL 30B-A3B Instruct
Alibaba Qwen3-VL
Multimodal MoE with only ~3B active parameters, so it stays responsive on Apple unified memory and survives CPU offload far better than a dense 30B. Q4 ~19GB — a 24GB card holds it outright.
36BConsumer GPUPro GPU19.5 GBmin VRAM+1.3%perplexity rise
Seed-OSS 36B Instruct
ByteDance Seed
Dense 36B with a native 512K context. The Q4_K_M file is 21.8 GB, so with cache and buffers it needs about 23.9 GB at 4K: tight on a 24 GB card, where Q3_K_M (19.5 GB) is the comfortable choice, or a 2×16GB split. The AWQ build is 21.1 GB, not the 4-bit rate the name suggests. Context is the real cost: 512K of KV cache is roughly 128GB on its own, so budget context first and weights second.
27BConsumer GPUPro GPU17.5 GBmin VRAM—perplexity rise
Qwen3.8 27B
Alibaba Qwen3.8
Hybrid attention: only 16 of its 64 layers keep a KV cache, which is why a 27B model holds a 262K context in about 16GB of cache rather than 64GB. Sizing it as a conventional stack overstates the cache fourfold — the estimates here count the 16 and match published measurements at 8K, 32K and 262K. Quality loss per level has not been published.
[30B-A3BConsumer GPUPro GPU19.2 GBmin VRAM—perplexity rise](https://quantized.uk/quant-hub/glm-4.7-flash/)
### GLM-4.7-Flash
Zhipu GLM-4.7
Zhipu's lightweight member of GLM-4.7: 30B total, about 3B active per token, MIT licence. Its attention is MLA, which caches one compressed 576-value vector per token per layer instead of full keys and values — so at 32K context the cache is under 2 GB where a conventional 47-layer model would need several times that. llama.cpp runs it as a DeepSeek-2-style model. Sizes below use the calculator's generic Q4_K_M/Q8_0 rates: no GGUF file size could be checked, and quality loss per level has not been published.
Running Nanbeige4.3 33B locally #
At Q4_K_M and 4K of context, Nanbeige4.3 33B needs about 22.6 GB — 19.7 GB of weights, 0.81 GB of KV cache and a 2.1 GB activation buffer. The smallest cards in this index that clear that comfortably are 32 GB ones — 4 of them, from the RTX 5090 up — and 26 of the 78 cards here do. These are calculated figures, not measurements: the estimate stops counting a card as comfortable at 88% of its VRAM, which is roughly the room a desktop session needs.
What longer context costs #
Going from 4K to 32K adds about 5.69 GB, taking the total to 28.9 GB. Weights do not move with context — only the KV cache does, and it grows linearly, so this is the number to watch when planning for long documents. The model's native window is 512K; holding all of it at Q4_K_M would need about 136 GB.
Which build to download #
This index only tracks Nanbeige4.3 33B in GGUF, across 3 levels (Q3_K_M, Q4_K_M, Q8_0). No per-level perplexity sweep has been published for this model, so the quality column is empty rather than estimated — within one model, more bits per weight is the only ordering the data supports. No throughput has been measured for this model on this site's hardware. GGUF runs on llama.cpp and Ollama across NVIDIA, AMD and Apple silicon; AWQ and GPTQ target vLLM on CUDA and ROCm; EXL2 is ExLlamaV2 — CUDA only, and now archived, so local use rather than serving.
Common questions #
- How much VRAM does Nanbeige4.3 33B need?
- About 22.6 GB at Q4_K_M with 4K of context and batch 1, rising to roughly 28.9 GB at 32K. That figure is weights plus KV cache plus a 10% activation buffer, calculated from the model's architecture rather than measured on a card.
- Will Nanbeige4.3 33B run on a 32GB GPU?
- Yes — at Q4_K_M and 4K context it needs about 22.6 GB, which leaves 9.4 GB spare on a RTX 5090, or on any of the other 32 GB cards here (4 in all). That is the smallest size in this index that clears it comfortably; 26 of 78 do.
- Which quantization of Nanbeige4.3 33B should I use?
- No published quality comparison exists for this model, so pick by footprint: Q4_K_M is the level most people run, and a higher bits-per-weight level is more faithful. All 3 levels in the index are Q3_K_M, Q4_K_M, Q8_0.
Where this model fits #
Sized at Q4_K_M with a 4K context window, smallest card first. Comfortable means the estimate uses at most 88% of the memory.
- Runs comfortably on
- RTX 509032GB · +9.4Arc Pro B70 32G32GB · +9.4Radeon AI PRO R9700 32G32GB · +9.4Instinct MI100 32G32GB · +9.4Mac M3 Pro 36G36GB · +4.4
- Tight but possible
- RTX 309024GBRadeon RX 7900 XTX24GBRTX 409024GB
- Ships in
- GGUF
- Try it yourself
- Size it in the calculatorGenerate the run command
How to actually run this #
Deployment guides for this model and this class of hardware.
This page's figures change when the model or the runtime does.Last updated 2026-10-26 RSS → /feed.xml