Granite 4.0 350M — Hardware Requirements & GPU Compatibility
Chat
Granite 4.0 350M is IBM's 350-million-parameter instruct model, the smallest in the Granite 4.0 Nano group, finetuned from Granite-4.0-350M-Base with supervised finetuning, reinforcement learning and model merging. It is meant for lightweight assistant tasks, with tool calling, fill-in-the-middle code completion and support for twelve languages including English, German, Japanese, Arabic and Chinese. Its config uses a hybrid architecture (GraniteMoeHybrid). At this size it runs on a CPU, a phone-class device or any small GPU without quantization, which suits edge and on-device use. The context window is 32,768 tokens. It is released under the Apache 2.0 license, permitting unrestricted commercial and research use. The Hub repository was created in October 2025, and the card lists a release date of October 28, 2025. It is the smaller sibling of Granite 4.0 1B, and the card says users can fine-tune the Nano models to add languages beyond the supported list.
24.5K downloads
71 likesSpecifications
- Publisher
- IBM
- Family
- Granite
- Parameters
- 352M
- Architecture
- GraniteMoeHybridForCausalLM
- Context Length
- 32,768 tokens
- Vocabulary Size
- 100,352
- Release Date
- 2025-10-07
- License
- Apache 2.0
Get Started
HuggingFace
How Much VRAM Does Granite 4.0 350M Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| Q2_K | 3.40 | 0.5 GB | 1.4 GB | 0.15 GB | 2-bit quantization with K-quant improvements |
| Q3_K_S | 3.50 | 0.5 GB | 1.4 GB | 0.15 GB | 3-bit small quantization |
| Q3_K_M | 3.90 | 0.5 GB | 1.4 GB | 0.17 GB | 3-bit medium quantization |
| Q4_0 | 4.00 | 0.5 GB | 1.4 GB | 0.18 GB | 4-bit legacy quantization |
| Q4_K_M | 4.80 | 0.6 GB | 1.4 GB | 0.21 GB | 4-bit medium quantization — most popular sweet spot |
| Q5_K_M | 5.70 | 0.6 GB | 1.5 GB | 0.25 GB | 5-bit medium quantization — good quality/size tradeoff |
| Q6_K | 6.60 | 0.7 GB | 1.5 GB | 0.29 GB | 6-bit quantization, very good quality |
| Q8_0 | 8.00 | 0.7 GB | 1.6 GB | 0.35 GB | 8-bit quantization, near-lossless |
Which GPUs Can Run Granite 4.0 350M?
Q4_K_M · 0.6 GBGranite 4.0 350M (Q4_K_M) requires 0.6 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 1+ GB is recommended. Using the full 33K context window can add up to 0.9 GB, bringing total usage to 1.4 GB. 52 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti.
Runs great
— Plenty of headroomWhich Devices Can Run Granite 4.0 350M?
Q4_K_M · 0.6 GB59 devices with unified memory can run Granite 4.0 350M, including NVIDIA DGX H100, NVIDIA DGX A100 640GB.
Runs great
— Plenty of headroomWhere to Download Granite 4.0 350M
Community quantizations of this model — GGUF for llama.cpp, Ollama, and LM Studio, plus AWQ/MLX variants where available.
Benchmarks
Benchmark details →Related Models
Frequently Asked Questions
- How much VRAM does Granite 4.0 350M need?
Granite 4.0 350M requires 0.6 GB of VRAM at Q4_K_M, or 1.1 GB at BF16. Full 33K context adds up to 0.9 GB (1.4 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 352M × 4.8 bits ÷ 8 = 0.2 GB
KV Cache + Overhead ≈ 0.4 GB (at 2K context + ~0.3 GB framework)
KV Cache + Overhead ≈ 1.3 GB (at full 33K context)
VRAM usage by quantization
Q4_K_M0.6 GBQ4_K_M + full context1.4 GB- What's the best quantization for Granite 4.0 350M?
For Granite 4.0 350M, Q4_K_M (0.6 GB) offers the best balance of quality and VRAM usage. Q5_0 (0.6 GB) provides better quality if you have the VRAM. The smallest option is IQ3_XXS at 0.5 GB.
VRAM requirement by quantization
IQ3_XXS0.5 GBQ4_00.5 GBIQ4_NL0.6 GBQ4_K_M ★0.6 GBQ5_K_S0.6 GBBF161.1 GB★ Recommended — best balance of quality and VRAM usage.
- Can I run Granite 4.0 350M on a Mac?
Granite 4.0 350M requires at least 0.5 GB at IQ3_XXS, which exceeds the unified memory of most consumer Macs. You would need a Mac Studio or Mac Pro with a high-memory configuration.
- Can I run Granite 4.0 350M locally?
Yes — Granite 4.0 350M can run locally on consumer hardware. At Q4_K_M quantization it needs 0.6 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is Granite 4.0 350M?
At Q4_K_M, Granite 4.0 350M can reach ~8421 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~1150 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 ÷ 0.6 × 0.65 = ~9123 tok/s
Estimated speed at Q4_K_M (0.6 GB)
~9123 tok/s~1150 tok/s~9123 tok/s~8421 tok/sReal-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
- What's the download size of Granite 4.0 350M?
At Q4_K_M, the download is about 0.21 GB. The full-precision BF16 version is 0.70 GB. The smallest option (IQ3_XXS) is 0.14 GB.
- Which GPUs can run Granite 4.0 350M?
52 consumer GPUs can run Granite 4.0 350M at Q4_K_M (0.6 GB). 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 350M?
59 devices with unified memory can run Granite 4.0 350M at Q4_K_M (0.6 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Apple iPhone 17 Pro, Asus ROG Flow Z13 (2025, 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.