Granite 4.0 Micro — Hardware Requirements & GPU Compatibility
ChatFunctions
Granite-4.0-Micro is IBM's 3-billion-parameter long-context instruct model, fine-tuned from Granite-4.0-Micro-Base on a mix of permissively licensed open datasets and IBM's own synthetic data, using supervised fine-tuning, reinforcement-learning alignment, and model merging. Unlike some of its Granite 4.0 siblings (H Micro, H Tiny MoE, H Small MoE), which mix in Mamba2 layers or mixture-of-experts routing, Granite-4.0-Micro is a plain decoder-only dense transformer with grouped-query attention, RoPE, SwiGLU, and RMSNorm. It targets enterprise use cases such as summarization, RAG, tool-calling, and code completion, and at 3B parameters it runs comfortably on a single consumer GPU, even quantized on a laptop. Context length is 131,072 tokens. It is released under the Apache 2.0 license, permitting unrestricted commercial and research use, and was published in October 2025.
Runs great on a 24 GB GPU
Runs from 8 GB (Q4_K_M · 64K)Best on 24 GB (Q4_K_M · ~261 tok/s · 128K)Macs from 8 GB (Q4_K_M · ~29 tok/s)
Apache 2.0
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Granite 4.0 Micro
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
ollama run hf.co/ibm-granite/granite-4.0-micro-GGUF:Q4_K_MQ4_K_M fits fully on the RTX 4090 · 20.3 GB free · 16K context
Download: granite-4.0-micro-Q4_K_M.gguf2.10 GBibm-granite
Other ways
each opens step-by-stepRun in cloud
Fits on RTX 3060 12GB (8 GB headroom) · Q4_K_M
- Generation speed
- ~93 tok/s
- generation speed
- Cost per 1M output tokens
- $0.25
- per 1M output tokens
How Much VRAM Does Granite 4.0 Micro Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 1.9 GB | 4.4 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 128K context · ~343 tok/s | |
| 2.0 GB | 4.5 GB | 3-bit small — 3-bit small quantization | Fits: 128K context · ~334 tok/s | |
| 2.1 GB | 4.7 GB | 3-bit medium — 3-bit medium quantization | Fits: 128K context · ~308 tok/s | |
| 2.2 GB | 4.7 GB | 4-bit · legacy — 4-bit legacy quantization | Fits: 128K context · ~302 tok/s | |
| 2.5 GB | 5 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 128K context · ~261 tok/s | |
| 2.9 GB | 5.4 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Fits: 128K context · ~227 tok/s | |
| 3.3 GB | 5.8 GB | 6-bit, very good — 6-bit quantization, very good quality | Fits: 128K context · ~200 tok/s | |
| 3.9 GB | 6.4 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Fits: 128K context · ~169 tok/s |
Selected Q4_K_M
4-bit medium quantization3.7GBat 16K context
- Weights
- 2.0 GB
- KV cache
- 1.3 GB
- Overhead
- 0.3 GB
Fits RTX 4090 (reference) · 20.3 GB free · ~261 tok/s
≥50 tok/s needs ≥193 GB/s · RTX 5090 ~464, RTX 3090 Ti ~261
Which Hardware Can Run Granite 4.0 Micro?
Q4_K_M · 2.5 GBGranite 4.0 Micro (Q4_K_M) requires 2.5 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 4+ GB is recommended. Using the full 131K context window can add up to 10.6 GB, bringing total usage to 13.1 GB. 52 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti. 163 devices with unified memory can run Granite 4.0 Micro, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB).
≤ 11 GB
≤ 11 GB tier →fits · 4.3 GB free · up to 64K ctx
16 GPUs · 1 Mac class · 2 device families
RTX 3080~197Simulate output on RTX 3080RTX 2080 Ti~160Simulate output on RTX 2080 TiRTX 3070 Ti~158Simulate output on RTX 3070 TiM3*8 GB~29Simulate output on iMac 24" M3 (8 GB)Orin Nano~26Simulate output on NVIDIA Jetson Orin Nano 8GB (Super)iPhone~19Simulate output on iPhone 1720 GB
fits · 16.3 GB free · up to 128K ctx
1 GPU
RX 7900 XT~191Simulate output on RX 7900 XT32–40 GB
fits · 28.3 GB free · up to 128K ctx
1 GPU · 3 Mac classes · 4 device families
RTX 5090~464Simulate output on RTX 5090M5 Max36 GB~128Simulate output on Mac Studio M5 Max (36 GB)M532 GB~43Simulate output on MacBook Air 13" M5 (32 GB)Orin~53Simulate output on NVIDIA Jetson AGX Orin 32GBSnapdragon X Elite (X1E-80-100)~35Simulate output on Snapdragon X Elite Copilot+ PC
By VRAM tier · typical GPUs and Macs at Q4_K_M
How good is Granite 4.0 Micro?
llmrun Score blends public benchmark boards into one number; 100 = the average open model.
Not enough data yet for an llmrun Score.
A Score needs results on enough independent boards. How the Composite Score is built
Benchmarks
Here are all 2 boards that have results for this model, sorted by their weight in the llmrun Score; badges show who ran the test.
- GPQA Diamond
Epoch AI · Oct 2025Independent
Independent28.3%
#84/98 open
- AIME 2024/2025
Epoch AI · Oct 2025Independent
Independent2.8%
#72/86 open
Alternatives to Granite 4.0 Micro
Granite family · 26 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Deepseek Vl2 TinyDeepSeek | — | 9.4 GB | ~152 | same size |
| Llama 3.2 3B InstructMeta | — | 4.1 GB | ~309 | same size |
| AI21 Jamba Reasoning 3BAI21 Labs | — | 2.3 GB | ~291 | same size |
| Sarashina2.2 3B Instruct v0.1sbintuitions | — | 3.7 GB | ~247 | same size |
| PowerMoE 3Bibm-research | — | 2.6 GB | ~354 | same size |
| MiniMax M3 EAGLE3Inferact | — | 2.7 GB | ~284 | same size |
Specifications
- Publisher
- IBM
- Family
- Granite
- Parameters
- 3.4B
- Architecture
- GraniteMoeHybridForCausalLM
- Context Length
- 131,072 tokens
- Vocabulary Size
- 100,352
- Release Date
- 2025-09-16
- License
- Apache 2.0
Frequently Asked Questions
How much VRAM does Granite 4.0 Micro need?
Granite 4.0 Micro requires 2.5 GB of VRAM at Q4_K_M, or 7.3 GB at BF16. Full 131K context adds up to 10.6 GB (13.1 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 3.4B × 4.8 bits ÷ 8 = 2 GB
KV Cache + Overhead ≈ 0.5 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 131K context)
VRAM usage by quantization
What's the best quantization for Granite 4.0 Micro?
For Granite 4.0 Micro, Q4_K_M (2.5 GB) offers the best balance of quality and VRAM usage. Q4_K_L (2.5 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 1.4 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Granite 4.0 Micro on a Mac?
Yes — MacBook Air 13" M3 (8 GB) and 145 other Macs can run Granite 4.0 Micro. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 2.5 GB of usable unified memory (RAM minus macOS overhead).
Can I run Granite 4.0 Micro locally?
Yes — Granite 4.0 Micro can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Granite 4.0 Micro?
At Q4_K_M, Granite 4.0 Micro can reach ~1912 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~261 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 ÷ 2.51 × 0.65 = ~2072 tok/s
Estimated speed at Q4_K_M (2.5 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Granite 4.0 Micro?
At Q4_K_M, the download is about 2.04 GB. The full-precision BF16 version is 6.81 GB. The smallest option (IQ2_XXS) is 0.94 GB.
Which GPUs can run Granite 4.0 Micro?
52 consumer GPUs can run Granite 4.0 Micro at Q4_K_M. Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6800, AMD Radeon RX 6800 XT. 52 GPUs have plenty of headroom for comfortable inference.
Which devices can run Granite 4.0 Micro?
163 devices with unified memory can run Granite 4.0 Micro at Q4_K_M (2.5 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.