alt.freeweights.granite / granite-3.2-8b-instruct-preview
From: ibm-granite <huggingface.co/ibm-granite>Newsgroups: alt.freeweights.graniteSubject:

granite-3.2-8b-instruct-preview

Date: 2026-11-15Commit: 191a764Signed: 2DC28CC7B26A283B ✓

Text generation model with 8.2B parameters. Safetensors weights, 16.3 GB in all. Published by ibm-granite under Apache-2.0.

8.2B params16.3GBgranite-3.2-8b-instruct-previewsafetensorspermissive
FactsParams8.2BSize16.3 GBFormatssafetensorsTasktext generationLicenseApache-2.0From sourcepermissive
$ fw get ibm-granite/granite-3.2-8b-instruct-preview
# manifest sha256 13637371fdea… · 16.3 GB · 12 files
$ fw verify ./granite-3.2-8b-instruct-preview ibm-granite/granite-3.2-8b-instruct-preview
From sourceHosted (signed)ibm-granite/granite-3.2-8b-instruct-previewLast check: 2026-11-15 03:32 UTC

Get help: custom fine-tune or private deploy. We review your setup before we quote.

Fine-tune of granite-3.1-8b-instruct.

Works with

Overview

Size
16.3 GB
Weights
16.3 GB (safetensors)
Parameters
8.2B
Formats
safetensors
Task
text generation
License
Apache-2.0 (permissive)
Files
12, each with a sha256
Publisher
ibm-granite
Added
2026-11-15
Section
alt.freeweights.granite
Lineage

Original release from the publisher.

Base model: ibm-granite/granite-3.1-8b-instruct

Where these bytes come from
Publisher at the pinned commit→You

Sources are tried in order. A file counts only when its size and sha256 match the manifest. Sources are checked when you ask for them, not on a timer.

Sources
BadgeHostLast checkResult
From sourcehuggingface.co2026-11-15 03:32 UTCpassed

Model card

From the publisher at commit 191a764. Shown as plain text with links; images open at the publisher.

Granite-3.2-8B-Instruct-Preview#

Model Summary: Granite-3.2-8B-Instruct-Preview is an early release of an 8B long-context model fine-tuned for enhanced reasoning (thinking) capabilities. Built on top of Granite-3.1-8B-Instruct, it has been trained using a mix of permissively licensed open-source datasets and internally generated synthetic data designed for reasoning tasks. The model allows controllability of its thinking capability, ensuring it is applied only when required.

Supported Languages: English, German, Spanish, French, Japanese, Portuguese, Arabic, Czech, Italian, Korean, Dutch, and Chinese. However, users may finetune this Granite model for languages beyond these 12 languages.

Intended Use: The model is designed to respond to general instructions and can be used to build AI assistants for multiple domains, including business applications.

Capabilities

  • Thinking
  • Summarization
  • Text classification
  • Text extraction
  • Question-answering
  • Retrieval Augmented Generation (RAG)
  • Code related tasks
  • Function-calling tasks
  • Multilingual dialog use cases
  • Long-context tasks including long document/meeting summarization, long document QA, etc.

Generation: This is a simple example of how to use Granite-3.2-8B-Instruct-Preview model.

Install the following libraries:

pip install torch torchvision torchaudio
pip install accelerate
pip install transformers

Then, copy the snippet from the section that is relevant for your use case.

from transformers import AutoModelForCausalLM, AutoTokenizer, set_seed
import torch

model_path="ibm-granite/granite-3.2-8b-instruct-preview"
device="cuda"
model = AutoModelForCausalLM.from_pretrained(
        model_path,
        device_map=device,
        torch_dtype=torch.bfloat16,
    )
tokenizer = AutoTokenizer.from_pretrained(
        model_path
)

conv = [{"role": "user", "content":"How do I develop a skill?"}]

input_ids = tokenizer.apply_chat_template(conv, return_tensors="pt", thinking=True, return_dict=True, add_generation_prompt=True).to(device)

set_seed(42)
output = model.generate(
    **input_ids,
    max_new_tokens=8192,
)

prediction = tokenizer.decode(output[0, input_ids["input_ids"].shape[1]:], skip_special_tokens=True)
print(prediction)

Example Outputs

  • thinking=True
Here is my thought process:

1. Understand the user's intent: The user is seeking guidance on how to develop a new skill. 
2. Break down the process: Skills development involves several stages - identification, planning, practice, feedback, and refinement. 
3. Consider different aspects: Discuss various methods, mindset, and resources needed for effective skill acquisition. 
4. Provide actionable steps: Offer a structured approach that the user can follow. 

Here is my response:

Developing a new skill is a systematic process that involves several key steps and a dedicated mindset. Here's a comprehensive guide to help you effectively cultivate a new skill:

### 1. **Identify the Skill**
   - **Clarity of Purpose**: Clearly define what skill you want to acquire. Be specific about why you want this skill and how it aligns with your goals or interests.
   - **Research**: Understand the basics of the skill. Read about it, watch tutorials, or talk to people who are already proficient. This initial research will give you a realistic understanding of what the skill entails.

### 2. **Set Clear Goals**
   - **SMART Goals**: Make sure your goals are Specific, Measurable, Achievable, Relevant, and Time-bound. For example, instead of saying "I want to learn programming," set a goal like "I will complete a beginner's Python course within three months."
   - **Break Down**: If the skill is complex, break it down into smaller, manageable milestones. 

### 3. **Create a Learning Plan**
   - **Resources**: Identify the best resources for learning. These could be online courses, books, mentors, workshops, or practical projects.
   - **Schedule**: Dedicate time in your routine for skill practice. Consistency is crucial in skill development. Treat your learning time as non-negotiable.

### 4. **Start with the Basics**
   - **Foundational Knowledge**: Begin by grasping the fundamentals. Understanding the basics thoroughly sets a strong foundation for more advanced learning.
   - **Practice Regularly**: Regular, focused practice is essential. Even 15-30 minutes a day can yield significant results over time.

### 5. **Seek Feedback and Guidance**
   - **Mentorship**: If possible, find a mentor or join a community of learners. Feedback from experienced individuals can accelerate your learning curve.
   - **Peer Learning**: Engage with others learning the same skill. Peer groups provide support, motivation, and different perspectives.

### 6. **Apply the Skill in Real-World Scenarios**
   - **Practical Application**: Look for opportunities to apply what you've learned. This could be through projects, internships, or volunteering. Practical application solidifies your understanding and builds confidence.
   - **Iterate Based on Feedback**: After applying your skill, seek and incorporate feedback to refine and improve.

### 7. **Maintain a Growth Mindset**
   - **Embrace Challenges**: Understand that setbacks and failures are part of the learning process. Maintain a positive attitude and view them as opportunities to learn.
   - **Patience and Persistence**: Skill development takes time. Stay patient and persistent, even when progress seems slow.

### 8. **Track and Reflect**
   - **Progress Tracking**: Regularly review your progress. This could be through keeping a learning journal or using apps designed for skill tracking.
   - **Reflect and Adjust**: Periodically reflect on your learning process. What's working? What's not? Adjust your plan accordingly.

### 9. **Stay Motivated**
   - **Reward System**: Set small rewards for achieving milestones to keep yourself motivated.
   - **Stay Curious**: Keep your interest piqued by exploring related topics or advanced aspects of the skill.

By following these steps and maintaining a disciplined, growth-oriented approach, you can effectively develop any new skill. Remember, the key is consistent effort and a willingness to learn from both successes and failures. 

Building skills is an ongoing journey, not a destination.
  • thinking=False
Developing a skill involves a combination of learning, practice, and often, feedback. Here's a step-by-step guide to help you develop a new skill:

1. **Identify the Skill**: Start by clearly defining what skill you want to develop. Be specific. Instead of saying "I want to learn to code," specify a programming language like Python or JavaScript.

2. **Research**: Learn about the basics of the skill. Read books, articles, watch tutorials, or take online courses. Websites like Coursera, Udemy, Khan Academy, and YouTube can be great resources.

3. **Set Clear Goals**: Break down your skill into smaller, manageable goals. For example, if you're learning a new language, your goals might be to learn basic grammar, build a simple sentence, have a basic conversation, etc.

4. **Create a Study Plan**: Allocate specific time each day or week for learning and practicing. Consistency is key in skill development.

5. **Practice**: Apply what you've learned. Practice makes permanent. If you're learning to code, write small programs. If it's a musical instrument, play regularly.

6. **Get Feedback**: Seek feedback from others who are more experienced. This could be a mentor, a tutor, or even online communities. Constructive criticism can help you identify areas for improvement.

7. **Review and Refine**: Regularly review what you've learned. Refine your skills based on feedback and your own observations.

8. **Apply in Real Life**: Try to use your new skill in real-life situations. This could be a project at work, a personal hobby, or volunteering.

9. **Be Patient and Persistent**: Skill development takes time. Don't get discouraged by slow progress or setbacks. Keep practicing and learning.

10. **Stay Motivated**: Keep your end goal in mind and celebrate small victories along the way to stay motivated.

Remember, everyone learns at their own pace, so don't compare your progress with others. The most important thing is that you're consistently moving forward.

Evaluation Results:

| Models | ArenaHard | Alpaca-Eval-2 | MMLU | PopQA | TruthfulQA | BigBenchHard | DROP | GSM8K | HumanEval | HumanEval+ | IFEval | AttaQ

| Llama-3.1-8B-Instruct | 36.43 | 27.22 | 69.15 | 28.79 | 52.79 | 72.66 | 61.48 | 83.24 | 85.32 | 80.15 | 79.10 | 83.43

| DeepSeek-R1-Distill-Llama-8B | 17.17 | 21.85 | 45.80 | 13.25 | 47.43 | 65.71 | 44.46 | 72.18 | 67.54 | 62.91 | 66.50 | 42.87

| Qwen-2.5-7B-Instruct | 25.44 | 30.34 | 74.30 | 18.12 | 63.06 | 70.40 | 54.71 | 84.46 | 93.35 | 89.91 | 74.90 | 81.90

| DeepSeek-R1-Distill-Qwen-7B | 10.36 | 15.35 | 50.72 | 9.94 | 47.14 | 65.04 | 42.76 | 78.47 | 79.89 | 78.43 | 59.10 | 42.45

| Granite-3.1-8B-Instruct | 37.58 | 27.87 | 66.84 | 28.84 | 65.92 | 68.10 | 50.78 | 79.08 | 88.82 | 84.62 | 71.20 | 85.73

| Granite-3.2-8B-Instruct-Preview | 55.23 | 61.16 | 66.93 | 28.08 | 66.37 | 65.60 | 50.73 | 83.09 | 89.47 | 86.88 | 73.57 | 85.99

Training Data: Overall, our training data is largely comprised of two key sources: (1) publicly available datasets with permissive license, (2) internal synthetically generated data targeted to enhance reasoning capabilites.

Infrastructure: We train Granite-3.2-8B-Instruct-Preview using IBM's super computing cluster, Blue Vela, which is outfitted with NVIDIA H100 GPUs. This cluster provides a scalable and efficient infrastructure for training our models over thousands of GPUs.

Ethical Considerations and Limitations: Granite-3.2-8B-Instruct-Preview builds upon Granite-3.1-8B-Instruct, leveraging both permissively licensed open-source and select proprietary data for enhanced performance. Since it inherits its foundation from the previous model, all ethical considerations and limitations applicable to Granite-3.1-8B-Instruct remain relevant.

Resources

Files

12 files · 16.3 GBWeights 16.3 GB (safetensors). Sizes and sha256 from the signed manifest.
Hide all 12 files with their sha256
FileSizesha256
.gitattributes1.52 KB11ad7efa24975ee4b0c3c3a38ed18737f0658a5f75a0a96787b576a78a023361
README.md19.3 KB50e00a8e56811dccb1fab1601f65dd5150b274a807d9a8035707ed4e51a78692
config.json856 Ba4a00d0d77e4a3316cfd45de3aad188b8cd8f11ff92d8e2cee51defadea93c75
generation_config.json132 B47ecbf011f713df5f0c8ecd10db3cd465380aea31ecf4c7c4b9bd3e7dbaa9d92
model-00001-of-00004.safetensors4.97 GBfe715006961601b91a7fb81340fbc094993515d0e5d43441982d6d9652dbee51
model-00002-of-00004.safetensors4.99 GB0c54831a14341386043e3a054fd75edd07e4bdb3793e7544b0842b9748f7b226
model-00003-of-00004.safetensors4.97 GB587600b0d79f7ecb8a85c7dc980cfd0b8b7a913d1d0174c6192e429a6fb08f8d
model-00004-of-00004.safetensors1.41 GB4649e6c9e75004c5537b0b3080ccd90ada10c90d5f6e84c49b80c314e4b4d31c
model.safetensors.index.json29.8 KB32fab6382f4cba5428af1b28ed350b6b33d6a413863e952fc10cc286328324e6
special_tokens_map.json701 B596d752bf46f5cace1f6826b52ed7d913347a4eea0ecce8ab2f869471ca40369
tokenizer.json3.48 MBed944c0b3d71e2d651a5ba0f8e7b350faff30031c8e3353de70b12b9c1bf5654
tokenizer_config.json8.88 KBfdc617feda7907bddc63907c2e922b1cbd677925b2068403cecc29c349a4725c
VersionsEach time the pinned upstream commit or file list changed in this repo. Newest first.
Upstream commitDateContents
191a764d71 pinned now2026-11-15 03:32 UTC12 files · 16.3 GB

Use it

$ fw get ibm-granite/granite-3.2-8b-instruct-preview
# manifest sha256 13637371fdea… · 16.3 GB · 12 files
$ fw verify ./granite-3.2-8b-instruct-preview ibm-granite/granite-3.2-8b-instruct-preview
fw.pyChecks the signature and every sha256, tries each source in order, resumes.
$ python3 fw.py get https://freeweights.allternit.com/manifests/ibm-granite/granite-3.2-8b-instruct-preview.json --dir ./granite-3.2-8b-instruct-preview $ python3 fw.py verify ./granite-3.2-8b-instruct-preview ./ibm-granite__granite-3.2-8b-instruct-preview.json
hfThe Hugging Face CLI with one setting. Same repo id.
$ HF_ENDPOINT=https://freeweights.allternit.com hf download ibm-granite/granite-3.2-8b-instruct-preview --revision 191a764d71575f2dbe6460d1576e16661d7f6d6b --local-dir ./granite-3.2-8b-instruct-preview
transformersLoad the weights in Python.
import os os.environ["HF_ENDPOINT"] = "https://freeweights.allternit.com" from transformers import AutoModelForCausalLM, AutoTokenizer tok = AutoTokenizer.from_pretrained("ibm-granite/granite-3.2-8b-instruct-preview", revision="191a764d71575f2dbe6460d1576e16661d7f6d6b") model = AutoModelForCausalLM.from_pretrained("ibm-granite/granite-3.2-8b-instruct-preview", revision="191a764d71575f2dbe6460d1576e16661d7f6d6b", device_map="auto")
vLLMServe an OpenAI-compatible API.
$ HF_ENDPOINT=https://freeweights.allternit.com vllm serve ibm-granite/granite-3.2-8b-instruct-preview --revision 191a764d71575f2dbe6460d1576e16661d7f6d6b

License

Name
Apache License 2.0
Id
Apache-2.0
Class
Permissive
Redistribution
yes
Listing
Hosted (signed)
Decided
License as declared by the publisher. No person read it. 2026-11-15
Text in
README.md
Text sha256
50e00a8e56811dccb1fab1601f65dd5150b274a807d9a8035707ed4e51a78692
Terms gate
No extra terms beyond the license.

Read the license text

Licenses are checked against the license table. A custom license passes only after a person reads it.

Provenance

From: ibm-granite <huggingface.co/ibm-granite>Newsgroups: alt.freeweights.graniteSubject: granite-3.2-8b-instruct-previewDate: 2026-11-15Commit: 191a764Signed: 2DC28CC7B26A283B ✓
Upstream
ibm-granite/granite-3.2-8b-instruct-preview
Pinned commit
191a764d71575f2dbe6460d1576e16661d7f6d6b
Manifest
granite-3.2-8b-instruct-preview.json · .minisig
Manifest sha256
13637371fdea865c81be86f2389b3735c025e4c33d3caece36ecd58acfa0811c
Signature
minisign, Ed25519, key 2DC28CC7B26A283B
Created
2026-11-15 03:32 UTC
Generator
freeweights-manifest/0.1.0
Takedowns
None.

How to verify

$ curl -O https://freeweights.allternit.com/manifests/ibm-granite/granite-3.2-8b-instruct-preview.json $ curl -O https://freeweights.allternit.com/manifests/ibm-granite/granite-3.2-8b-instruct-preview.json.minisig $ minisign -Vm granite-3.2-8b-instruct-preview.json -P RWQ7KGqyx4zCLXtUFOqvfv8L3g+8fpKJ1FsuxPyOYMF+4KufbUxKq1Oq $ shasum -a 256 <file> # compare with the Files tab

A manifest whose signature fails must be rejected. A file whose size or sha256 does not match must be discarded. Verify a download

Model tree

Parent

ibm-granite/granite-3.1-8b-instruct · this model is a finetune of it

No model in the catalog names this one as its parent.

Get it
Download
$ HF_ENDPOINT=https://freeweights.allternit.com hf download ibm-granite/granite-3.2-8b-instruct-preview
Open in Freeweights Studio

Pull, verify and run it in the Allternit app on Desktop, web or phone.

Signed manifest
minisign, key 2DC28CC7B26A283B. granite-3.2-8b-instruct-preview.json. Provenance. Takedown history: none.
From: ibm-granite <huggingface.co/ibm-granite>Subject: GneissWeb.7B_ablation_model_on_350B_FineWeb.Edu.seed1
From: ibm-granite <huggingface.co/ibm-granite>Subject: GneissWeb.7B_ablation_model_on_350B_FineWeb.Edu.seed2
From: ibm-granite <huggingface.co/ibm-granite>Subject: GneissWeb.7B_ablation_model_on_350B_FineWeb.Edu.seed3
From: ibm-granite <huggingface.co/ibm-granite>Subject: GneissWeb.7B_ablation_model_on_350B_FineWeb.seed1
From: ibm-granite <huggingface.co/ibm-granite>Subject: GneissWeb.7B_ablation_model_on_350B_FineWeb.seed2
From: ibm-granite <huggingface.co/ibm-granite>Subject: GneissWeb.7B_ablation_model_on_350B_FineWeb.seed3