Skip to content
TrustList

TrustList summary

Granite 3.0 8B Instruct is IBM's instruction-following model in the Granite 3.0 generation, released on 21 October 2024 and published on Hugging Face under the Apache 2.0 licence. It has about 8 billion parameters (8.1 billion by the card's table) and was finetuned from Granite 3.0 8B Base by the Granite Team at IBM.

How it was tuned

The card lists the techniques as supervised finetuning, alignment with reinforcement learning and model merging, using a structured chat format. The data blends open instruction datasets with permissive licences, synthetic data collected internally to target particular skills, and a very small amount of human-curated material. Training ran on IBM's Blue Vela cluster of NVIDIA H100 GPUs, which the card says is powered entirely by renewable energy.

Architecture in the Granite 3.0 family

The model is a decoder-only dense transformer using grouped query attention, rotary position embeddings, SwiGLU MLPs and RMSNorm, with shared input and output embeddings. The card's comparison table sets the 8B model beside a 2B dense model and two Mixture-of-Experts models of 1B and 3B size. For the 8B it gives an embedding size of 4096, 40 layers, 32 attention heads, a sequence length of 4096 and 12 trillion training tokens.

Tasks and languages

IBM designed the model to answer general instructions and to underpin assistants across business domains. Its listed capabilities are summarization, text classification, text extraction, question answering, retrieval augmented generation, code tasks, function calling and multilingual dialog. Twelve languages are supported: Chinese and Japanese and Korean, Arabic, plus English and the major European languages German, Spanish, French, Portuguese, Czech, Italian and Dutch. Users may finetune it for others.

Stated limits

Instruction data is mostly English, so non-English performance may fall short, and the card suggests a few examples in the prompt to help. It also warns of inaccurate, biased or unsafe output and urges safety testing before deployment.

Summary written by TrustList from the business's own published information, Nov 2026.

Benchmarks & AI stats

BenchmarkOfficialCommunity avg
No benchmark scores yet. Be the first to add one.

“Official” values are editor-approved and feed the ranking. “Community avg” is the mean of member submissions (shown for transparency; it never affects the ranking until an editor approves a value).

Sign in to add a benchmark score for this model.

Granite 3.0 8B launches

Request a demo or quote from Granite 3.0 8B

Protected by reCAPTCHA - Google Privacy Policy and Terms apply.

By sending, you agree we may share your request and contact details with the provider once you confirm your email.

Reviews

Write the first review of Granite 3.0 8B

Used it? Your experience helps other buyers decide.

Write a review

Questions & answers

No questions yet. Be the first to ask about Granite 3.0 8B.

Granite 3.0 8B alternatives

Similar listings buyers compare against this one.

Microsoft AI's text-to-speech model for keeping one voice across 23 languages, with voice cloning from short audio

Perplexity's open-weight decision model that picks among given options for routing and checks in AI systems

Black Forest Labs' image generation and editing model with bounding-box layouts, multiple reference images and 4K output

Anthropic's smallest and fastest Claude 5.5 model, available on its own API and three cloud platforms

Microsoft AI's real-time speech-to-text model for voice agents, transcribing 60 languages with continuous language detection

JetBrains' open 12B mixture-of-experts coding model with 2.5B parameters active per token under Apache 2.0

Open-source 2B decision model from AWS's Strands Agents project that picks options and returns scores with a reliability figure

Google's open multimodal embedding model that maps text, images, audio and video into one vector space, under Apache 2.0