Granite Switch 4.1 8B — Hardware Requirements & GPU Compatibility
ChatFunctions
Granite Switch 4.1 8B is IBM's modular model that combines a Granite 4.1 8B base with two adapters and a speech encoder, so a single endpoint can chat, make structured decisions, check safety, and transcribe audio. A decision adapter returns probabilities over labeled options in one forward pass, and a Guardian adapter detects risks such as harm, jailbreaking, and bias. Adapters switch on through control tokens in the chat template. With roughly 8.8 billion parameters, a Q4 build fits a 8 to 12 GB GPU, though the composed format is mainly served through Transformers and vLLM. The context window is 131,072 tokens. It is released under the Apache 2.0 license and was published on October 14, 2026. Speech input uses an embedded 470 million parameter English transcriber, so audio goes through the same API call as text. IBM also offers a 30B version with the same configuration.
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Granite Switch 4.1 8B
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
vllm serve ibm-granite/granite-switch-4.1-8b --max-model-len 16384vLLM repo with its own quantisation
Download: Weights on Hugging Face
Other ways
each opens step-by-stepRun in cloud
Fits on RTX A6000 (48 GB) (27 GB headroom) · BF16
- Generation speed
- ~27 tok/s
- generation speed
- Cost per 1M output tokens
- $3.35
- per 1M output tokens
How Much VRAM Does Granite Switch 4.1 8B Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 18.2 GB | 23.5 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Fits: 32K context · ~36 tok/s |
Selected BF16
Brain floating point 1620.7GBat 16K context
- Weights
- 17.6 GB
- KV cache
- 2.8 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 3.3 GB free · ~36 tok/s
≥50 tok/s needs ≥1.40 TB/s · RTX 5090 ~64
Which Hardware Can Run Granite Switch 4.1 8B?
BF16 · 18.2 GBGranite Switch 4.1 8B (BF16) requires 18.2 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 24+ GB is recommended. Using the full 131K context window can add up to 22.2 GB, bringing total usage to 40.4 GB. 8 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti. 125 devices with unified memory can run Granite Switch 4.1 8B, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, Mac Mini M5 Pro (24 GB).
20 GB
tight · up to 8K ctx
1 GPU
RX 7900 XT~26Simulate output on RX 7900 XT32–40 GB
fits · 11.3 GB free · up to 64K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~64Simulate output on RTX 5090M5 Max36 GB~18Simulate output on Mac Studio M5 Max (36 GB)M532 GB~6Simulate output on MacBook Air 13" M5 (32 GB)Orin~7Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~5Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at BF16
Alternatives to Granite Switch 4.1 8B
Granite family · 26 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| ALoDLM 8BAmazon | — | 7.6 GB | ~119 | same size |
| Llm Jp 4.1 8B Thinkingllm-jp | — | 7.6 GB | ~115 | same size |
| Apertus V1.5 8Bswiss-ai | — | 8.5 GB | ~111 | same size |
| Ornith 1.0 9BDeep Reinforce | — | 6.5 GB | ~106 | same size |
| LFM2.5 8B A1BLiquid AI | — | 5.6 GB | ~335 | same size |
| MiniCPM O 4 5OpenBMB | — | 8.3 GB | ~105 | same size |
Specifications
- Publisher
- IBM
- Family
- Granite
- Parameters
- 8.8B
- Architecture
- GraniteSwitchForCausalLM
- Context Length
- 131,072 tokens
- Vocabulary Size
- 100,355
- Release Date
- 2026-10-14
- License
- Apache 2.0
Frequently Asked Questions
How much VRAM does Granite Switch 4.1 8B need?
Granite Switch 4.1 8B requires 18.2 GB of VRAM at BF16. Full 131K context adds up to 22.2 GB (40.4 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 8.8B × 16 bits ÷ 8 = 17.6 GB
KV Cache + Overhead ≈ 0.6 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 ≈ 22.8 GB (at full 131K context)
VRAM usage by quantization
Can I run Granite Switch 4.1 8B on a Mac?
Yes — Mac Mini M2 (24 GB) and 106 other Macs can run Granite Switch 4.1 8B. Apple Silicon uses unified memory, so the model shares RAM with the system. At BF16 you need at least 18.2 GB of usable unified memory (RAM minus macOS overhead).
Can I run Granite Switch 4.1 8B locally?
Yes — Granite Switch 4.1 8B can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Granite Switch 4.1 8B?
At BF16, Granite Switch 4.1 8B can reach ~264 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~36 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 18.2 × 0.65 = ~286 tok/s
Estimated speed at BF16 (18.2 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Granite Switch 4.1 8B?
At BF16, the download is about 17.55 GB.
Which GPUs can run Granite Switch 4.1 8B?
8 consumer GPUs can run Granite Switch 4.1 8B at BF16. Top options include NVIDIA GeForce RTX 5090, AMD Radeon RX 7900 XT, AMD Radeon RX 7900 XTX. 1 GPU have plenty of headroom for comfortable inference.
Which devices can run Granite Switch 4.1 8B?
125 devices with unified memory can run Granite Switch 4.1 8B at BF16 (18.2 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Asus ROG Flow Z13 (2025, Ryzen AI Max+ 395, 128 GB), Beelink GTR9 Pro (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.