Whittle Next 27B A3B — Hardware Requirements & GPU Compatibility
ChatReasoningCodeMath
Whittle Next 27B A3B is a 27.6B-parameter open language model from logic65. It supports a context window of up to 262,144 tokens. At Q4_K_M it needs about 16.92 GB of VRAM — see which GPUs and Macs can run it below.
Runs great on a 24 GB GPU
Runs from 16 GB (Q3_K_M · 32K)Best on 24 GB (Q4_K_M · ~129 tok/s · 128K)Macs from 18 GB (Q3_K_M · ~28 tok/s)
Apache 2.0
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Whittle Next 27B A3B
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
ollama run hf.co/logic65/Whittle-Next-27B-A3B-GGUF:Q4_K_MQ4_K_M fits fully on the RTX 4090 · 6.5 GB free · 16K context
Download: Whittle-Next-27B-A3B-v4.4-Q4_K_M.gguf16.80 GBlogic65
Ways to runOther ways
each opens step-by-stepRun in cloud
Runs on RTX A5000 (24 GB) (6 GB headroom) · Q4_K_M
- Generation speed
- ~105 tok/s
- generation speed
- Cost per 1M output tokens
- $0.42
- per 1M output tokens
How Much VRAM Does Whittle Next 27B A3B Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 12.1 GB | 13.4 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 256K context · ~162 tok/s | |
| 13.8 GB | 15.1 GB | 3-bit medium — 3-bit medium quantization | Fits: 128K context · ~149 tok/s | |
| 16.9 GB | 18.2 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 128K context · ~129 tok/s | |
| 20.0 GB | 21.3 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Fits: 64K context · ~114 tok/s | |
| 23.1 GB | 24.4 GB | 6-bit, very good — 6-bit quantization, very good quality | Fits: 16K context · ~102 tok/s | |
| 27.9 GB | 29.2 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Does not fit: needs 28.5 GB | |
| 55.5 GB | 56.8 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Does not fit: needs 56.1 GB |
Selected Q4_K_M
4-bit medium quantization17.5GBat 16K context
- Weights
- 16.5 GB
- KV cache
- 0.7 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 6.5 GB free · ~129 tok/s
≥50 tok/s needs ≥314 GB/s · RTX 5090 ~188, RTX 3090 Ti ~129
Which Hardware Can Run Whittle Next 27B A3B?
Q4_K_M · 16.9 GBWhittle Next 27B A3B (Q4_K_M) requires 16.9 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 22+ GB is recommended. Using the full 262K context window can add up to 10.7 GB, bringing total usage to 27.6 GB. 8 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti. 122 devices with unified memory can run Whittle Next 27B A3B, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Mini M5 Pro (24 GB).
20 GB
fits · 2.5 GB free · up to 64K ctx
1 GPU
RX 7900 XT~64Simulate output on RX 7900 XT32–40 GB
fits · 14.5 GB free · up to 256K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~188Simulate output on RTX 5090M5 Max36 GB~52Simulate output on Mac Studio M5 Max (36 GB)M532 GB~24Simulate output on MacBook Air 13" M5 (32 GB)Orin~34Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~22Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at Q4_K_M
Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Qwen3.8 27BAlibaba | 122.0 | 17.9 GB | ~38 | same size |
| Qwen3.6 27BAlibaba | 111.2 | 17.9 GB | ~38 | same size |
| Qwen3.5 27BAlibaba | 103.8 | 17.9 GB | ~38 | same size |
| Gemma 3 27B ITGoogle | 79.4 | 21.7 GB | ~36 | same size |
| Gemma 4 26B A4BGoogle | — | 16.5 GB | ~171 | same size |
| ERNIE 4.5 VL 28B A3B PTBaidu | — | 60 GB | —needs > 24 GB | same size |
Fine-tunes & variants(1)
Specifications
- Publisher
- logic65
- Parameters
- 27.6B
- Active parameters
- 5.9B per token
- Architecture
- Qwen4ExpForCausalLM
- Context Length
- 262,144 tokens
- Vocabulary Size
- 248,320
- Release Date
- 2026-09-03
- License
- Apache 2.0
Frequently Asked Questions
How much VRAM does Whittle Next 27B A3B need?
Whittle Next 27B A3B requires 16.9 GB of VRAM at Q4_K_M, or 55.5 GB at BF16. Full 262K context adds up to 10.7 GB (27.6 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 27.6B × 4.8 bits ÷ 8 = 16.5 GB
KV Cache + Overhead ≈ 0.4 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 11.1 GB (at full 262K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 4090 run Whittle Next 27B A3B?
Yes, at Q6_K (23.1 GB) or lower. Higher quantizations like Q8_0 (27.9 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.
What's the best quantization for Whittle Next 27B A3B?
For Whittle Next 27B A3B, Q4_K_M (16.9 GB) offers the best balance of quality and VRAM usage. Q5_K_M (20.0 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 12.1 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Whittle Next 27B A3B on a Mac?
Yes — MacBook Pro 14" M3 Pro (18 GB) and 108 other Macs can run Whittle Next 27B A3B, the smallest at Q3_K_M. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 16.9 GB of usable unified memory (RAM minus macOS overhead), from Mac Mini M5 Pro (24 GB).
Can I run Whittle Next 27B A3B locally?
Yes — Whittle Next 27B A3B can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Whittle Next 27B A3B?
At Q4_K_M, Whittle Next 27B A3B can reach ~114 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~129 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)
Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.
Example: NVIDIA B200 → 3.6 GB active ÷ (8000 × 0.65) = 0.70 ms, plus 40 layers × 0.055 ms = 2.20 ms, so 1000 ÷ 2.90 ms = ~345 tok/s
Estimated speed at Q4_K_M (16.9 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Whittle Next 27B A3B?
At Q4_K_M, the download is about 16.54 GB. The full-precision BF16 version is 55.14 GB. The smallest option (Q2_K) is 11.72 GB.
Which GPUs can run Whittle Next 27B A3B?
8 consumer GPUs can run Whittle Next 27B A3B at Q4_K_M. Top options include NVIDIA GeForce RTX 5090, AMD Radeon RX 7900 XT, AMD Radeon RX 7900 XTX. 1 GPU has plenty of headroom for comfortable inference.
Which devices can run Whittle Next 27B A3B?
122 devices with unified memory can run Whittle Next 27B A3B at Q4_K_M (16.9 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.