granite 4.2 3b
3.7B parameters · 128k context · Apache 2.0 · released 2026-08-25
chatcodingagenticreasoning
In short
granite 4.2 3b is an open model from IBM with 3.7B parameters. At Q4_K_M, granite 4.2 3b needs about 4 GB of memory at 8k context and 8 GB at 64k, counting the weights, the KV cache and runtime buffers. The cheapest hardware in Canada that runs granite 4.2 3b fully in memory at 64k context is the NVIDIA GeForce RTX 3060 12GB, $497 at Amazon, at about 24 tok/s (estimated).
Updated . Speeds are estimates; how they are worked out.
What to run it on in Canada
Fully in memory at 64k context, priced at retailers today.
Cheapest
NVIDIA GeForce RTX 3060 12GB24tok/sest.Q8_0
$497
Best value
NVIDIA GeForce RTX 3080 10GB56tok/sest.Q6_K
$769
Fastest
NVIDIA RTX PRO 6000 Blackwell120tok/sest.Q8_0
$3,133
All hardware that runs it
Everything we price, best value first. Two cards split the model by layers.
Memory needed
Weights, KV cache and runtime buffers at each context length (KV cache at fp16).
| Quant | File | @8k | @32k | @64k | @128k |
|---|---|---|---|---|---|
| Q4_K_Mgguf | 2.2 GB | 3.5 GB | 5.5 GB | 8.2 GB | 13.6 GB |
| Q5_K_Mgguf | 2.5 GB | 3.9 GB | 5.9 GB | 8.6 GB | 13.9 GB |
| Q6_Kgguf | 2.9 GB | 4.3 GB | 6.3 GB | 9.0 GB | 14.4 GB |
| Q8_0gguf | 3.6 GB | 5.0 GB | 7.0 GB | 9.7 GB | 15.1 GB |
Run it
Copy a command for the runtime you use.
- Q4_K_M
- Ollama
ollama run hf.co/bartowski/granite-4.2-3b-GGUF:Q4_K_M - LM Studio
lms get bartowski/granite-4.2-3b-GGUF@q4_k_m - llama.cpp
llama-server -hf bartowski/granite-4.2-3b-GGUF:Q4_K_M
- Ollama
- Q5_K_M
- Ollama
ollama run hf.co/bartowski/granite-4.2-3b-GGUF:Q5_K_M - LM Studio
lms get bartowski/granite-4.2-3b-GGUF@q5_k_m - llama.cpp
llama-server -hf bartowski/granite-4.2-3b-GGUF:Q5_K_M
- Ollama
- Q6_K
- Ollama
ollama run hf.co/bartowski/granite-4.2-3b-GGUF:Q6_K - LM Studio
lms get bartowski/granite-4.2-3b-GGUF@q6_k - llama.cpp
llama-server -hf bartowski/granite-4.2-3b-GGUF:Q6_K
- Ollama
- Q8_0
- Ollama
ollama run hf.co/bartowski/granite-4.2-3b-GGUF:Q8_0 - LM Studio
lms get bartowski/granite-4.2-3b-GGUF@q8_0 - llama.cpp
llama-server -hf bartowski/granite-4.2-3b-GGUF:Q8_0
- Ollama
Notes
Native context is 128K, and the card also claims a long-context extension to 512K; its serving examples use 131072. The vLLM reasoning parser needs vLLM v0.20+ (or SGLang v0.5.18+), and the recommended sampling is temperature=1.0 and top_p=0.95.
Questions
- How much VRAM does granite 4.2 3b need?
- At Q4_K_M, granite 4.2 3b needs about 4 GB of memory at 8k context and 8 GB at 64k, counting the weights, the KV cache and runtime buffers.
- What is the best GPU for running granite 4.2 3b locally?
- The cheapest hardware in Canada that runs granite 4.2 3b fully in memory at 64k context is the NVIDIA GeForce RTX 3060 12GB, $497 at Amazon, at about 24 tok/s (estimated). The best value is the NVIDIA GeForce RTX 3080 10GB (about 56 tok/s for $769). The fastest we price is the NVIDIA RTX PRO 6000 Blackwell, at about 120 tok/s.
- Can I run granite 4.2 3b on a 16 GB GPU?
- Yes: at Q8_0 it fits fully in 16 GB at 64k context. On the NVIDIA GeForce RTX 5080 it runs at about 64 tok/s (estimated).
- Can I run granite 4.2 3b on a 24 GB GPU?
- Yes: at Q8_0 it fits fully in 24 GB at 64k context. On the NVIDIA GeForce RTX 3090 Ti it runs at about 68 tok/s (estimated).
- How do I run granite 4.2 3b locally?
- With Ollama: ollama run hf.co/bartowski/granite-4.2-3b-GGUF:Q4_K_M (the Q4_K_M build). LM Studio and llama.cpp commands are on this page.
- What licence is granite 4.2 3b released under?
- Apache 2.0, which allows commercial use.