Model library / ibm-granite
Granite-4.2-3B is a 3.7 billion-parameter decoder-only language model with built-in chain-of-thought reasoning, released by IBM under Apache 2.0.
Parameters in total
Context length in tokens, as published
Weights at the released precision
Licence declared on the repository
Overview
Granite-4.2-3B is a decoder-only dense transformer model from IBM's Granite family, aimed at enterprise and research applications. IBM describes the model as versatile, safe and efficient, providing built-in chain-of-thought reasoning for tasks such as mathematics, coding and multi-step logic. The 4.2 generation adds native reasoning capabilities to improve performance on complex problem-solving.
In practice the model has 3.7 billion parameters and runs in bfloat16 precision. It supports a 131 072-token context window, enabling long documents and multi-turn interactions. The architecture uses GraniteForCausalLM with grouped query attention, 40 heads and 8 KV heads, and no expert layers. It is released under the Apache 2.0 licence, allowing unrestricted commercial use.
Specification
| Repository | ibm-granite/granite-4.2-3b |
| Publisher | ibm-granite |
| Published | 7 August 2026 |
| Architecture | GraniteForCausalLM |
| Model type | granite |
| Parameters | 3,659,737,600 |
| Layers | 40 |
| Hidden size | 2,560 |
| Attention heads | 40, 8 key/value heads |
| Context length | 131,072 tokens |
| Vocabulary | 100,352 tokens |
| Weight format | bfloat16 |
| Licence | apache-2.0 |
| Base model | ibm-granite/granite-4.1-3b-base |
Memory
Parameter count multiplied by the bytes each format uses. The key-value cache comes on top.
| Precision | Weights | Notes |
|---|---|---|
| bf16 | 6.8 GiB | Full precision as released |
| fp8 | 3.4 GiB | 8-bit, near-lossless on most models |
| int4 | 1.7 GiB | 4-bit, smallest footprint |
The cache costs 80 KiB per token at 16-bit, so the full 131,072-token context of one request needs about 10.0 GiB. Concurrency multiplies that number, not the weights.
Hardware
At the released precision, with 90% of the card given to the server and 2.5 GiB kept for the runtime. The context column is what is left for the cache on that setup.
| GPU | Cards needed | Free for cache | Context it holds |
|---|---|---|---|
| RTX 4090 24GB | 1 | 12.3 GiB | the full 131,072 tokens |
| RTX 5090 32GB | 1 | 19.5 GiB | the full 131,072 tokens |
| L40S 48GB | 1 | 33.9 GiB | the full 131,072 tokens |
| A100 80GB | 1 | 62.7 GiB | the full 131,072 tokens |
| H100 80GB | 1 | 62.7 GiB | the full 131,072 tokens |
| RTX PRO 6000 96GB | 1 | 77.1 GiB | the full 131,072 tokens |
| H200 141GB | 1 | 117.6 GiB | the full 131,072 tokens |
Serving
A vLLM launch line for the shape above. Check the model card for a runtime the publisher recommends.
vllm serve ibm-granite/granite-4.2-3b \ --max-model-len 131072 \ --gpu-memory-utilization 0.90
Questions
The weights take about 7 GiB at the released precision. That fits on one RTX 4090 24GB, which leaves roughly 12 GiB for the key-value cache.
The repository declares apache-2.0. Read the licence text before commercial use: the name alone does not tell you what is allowed.
The configuration allows 131,072 tokens. Whether the whole window is usable depends on the memory left for the cache, which the hardware table works out per card.
This page is a reference, and Aiqre does not serve this model today. We deploy open-weight models on dedicated EU hardware on request, so if you want an OpenAI-compatible endpoint for it, email [email protected] with the model name and roughly what volume you expect.
Also in the library
Granite-4.2-30B is a 29.3 billion-parameter decoder-only language model from IBM, designed for reasoning tasks with built-in chai…
Granite-4.2-8B is an 8.8 billion-parameter decoder-only language model from IBM, designed for reasoning-intensive tasks.
MiniCPM5-2B is a 2-billion-parameter dense transformer aimed at on-device and resource-constrained deployments.
NeoHorse-1-4B is a 4.2B open-weight model published by TokenRhythm on Hugging Face. It uses the Qwen3_5ForCausalLM architecture w…
Hosting
Tell us the model and roughly what volume you expect. We reply with a price and an OpenAI-compatible endpoint, with your prompts never stored.