Granite 4.2 8B
Granite 4.2 8B is the mid-size model of IBM's Granite 4.2 family, released in August 2026 under Apache 2.0. It shares the 30B model's thinking, non-thinking and low-effort modes and OpenAI-style function calling. At about 9B parameters it fits a single 24 GB GPU at BF16.
Specifications
- Context window
- 128Kabout 98,000 words, or 200 pages
- Max output
- Not published
- Input price
- $0.06per 1M tokens
- Output price
- $0.25per 1M tokens
- Cached input
- $0.015per 1M tokens
- Weights
- OpenApache-2.0
- Parameters
- 8.8Bdense
- Input types
- Text
- Output types
- Text
- Reasoning mode
- Yes
- Tool calling
- Yes
- Released
- Aug 25, 2026Current
DeepInfra serverless price (131,072 context). Native context is 128K; IBM says it can be extended to 512K.
What it costs in practice
| Workload | Monthly cost at list price |
|---|---|
| Support chatbot30,000 conversations a month, 2,000 input and 400 output tokens each | $6.60 |
| Document summaries2,000 documents a month, 12,000 input and 800 output tokens each | $1.84 |
| Coding agent1,000 tasks a month, 60,000 input and 6,000 output tokens each | $5.10 |
Standard rates, no caching or batch discounts. Run your own numbers.
Strengths
- Fits a 24 GB GPU without quantization
- Apache 2.0 license
- Reasoning before tool calls
Trade-offs to weigh
- Text-only input
- Lower ceiling on complex tasks than larger open models
Where it fits
IBM positions Granite 4.2 8B for:
Where you can use it
- Hugging Face
- Ollama
- DeepInfra
Run it yourself
Released under Apache-2.0. Approximate memory needed just to load the weights:
| Precision | Memory for weights |
|---|---|
| 16-bit (BF16/FP16) | 18 GB |
| 8-bit (FP8/INT8) | 9 GB |
| 4-bit (Q4_K_M, about 4.85 bits) | 5 GB |
Add memory for the KV cache (it grows with context length and concurrent users) and 1-2 GB of runtime overhead. Size it precisely in the VRAM calculator.
Alternatives to consider
| Model | Tier | Context | Input / output per 1M | |
|---|---|---|---|---|
| Ministral 3 3BMistral AI | Compact | 256K | $0.10 / $0.10 | Compare |
| gpt-oss-20bOpenAI | Compact | 128K | $0.075 / $0.30 | Compare |
| NVIDIA Nemotron 3.5 Lightning 30B-A3BNVIDIA | Compact | 1M | $0.05 / $0.20 | Compare |
| Gemma 4 26B A4BGoogle | Compact | 256K | $0.07 / $0.34 | Compare |
Frequently asked questions
How much does Granite 4.2 8B cost?
Granite 4.2 8B lists at $0.06 per million input tokens and $0.25 per million output tokens, with cached input at $0.015 (checked October 10, 2026). DeepInfra serverless price (131,072 context). Native context is 128K; IBM says it can be extended to 512K. Confirm current rates on the provider's pricing page before you commit.
What is the context window of Granite 4.2 8B?
Granite 4.2 8B accepts up to 131,072 tokens of context (about 98,000 words, or 200 pages). Long prompts cost more and can be slower, so most apps send only what the task needs.
Is Granite 4.2 8B open source?
Granite 4.2 8B is an open-weight model released under the Apache-2.0 license. You can download the weights from Hugging Face (ibm-granite/granite-4.2-8b) and run or fine-tune them yourself, within the license terms. Open weights isn't the same as fully open source: training data and code may not be published.
Can I run Granite 4.2 8B locally?
Yes, if you have enough memory. The weights alone need about 5 GB at 4-bit quantization and about 18 GB at 16-bit, plus room for the KV cache. Use the VRAM calculator on LLModel.com to check a specific GPU or Mac.
What is Granite 4.2 8B best for?
IBM positions Granite 4.2 8B for enterprise, self-hosting, low latency, fine-tuning. It accepts text as input, has a reasoning (thinking) mode and supports tool calling. Test it on a sample of your own tasks before you choose it for production.
Sources
- huggingface.co/ibm-granite/granite-4.2-8b
- deepinfra.com/ibm-granite/granite-4.2-8b
- huggingface.co/ibm-granite/granite-4.2-8b/raw/main/config.json
- huggingface.co/api/models/ibm-granite/granite-4.2-8b
- www.ibm.com/granite
- ollama.com/library
Facts checked against these pages on October 10, 2026. Spot an error? Tell us.
Still choosing a model?
Send us your use case and constraints. We'll come back with a shortlist and an evaluation plan you can run on your own data.
- Free and vendor-neutral: no model provider pays for placement.
- Your details go only to partners you agree to.