IBM Research

Researched 16d ago

61 models across 10 families · Latest: Granite 4.1 3B (2026-04)

Creating reliable and adaptable AI solutions

CodingRAGAgentsLong contextVisionClassificationJSON / Tool use

IBM Research's portfolio covers 59 active models across 10 non-obsolete families, with task labels spanning coding, rag, and agents. Open a model detail page to compare provider routes and sourced benchmarks.
Portfolio context: 7 decision-task tags, 59 active tracked models, latest research stamp 2026-05-19.

Use this portfolio page for

Teams evaluating coding, rag, and agents across this lab's releases
Readers comparing families before locking a flagship SKU
59 tracked SKUs for migration and pricing follow-ups

Do not stop here for

Choosing a hosting provider without opening a model page for price ladders

Active models

Non-deprecated SKUs linked to this researcher

Active families

Non-obsolete families in coverage

Open catalog

40 OSS

0 open-weight (text match)

Decision task tags

Mapped to the site-wide task taxonomy

Latest dated release

2026-04-29

Granite 4.1 3B

Freshness

2026-05-19

Researched 16d ago

fresh

Release cadence

Showing 5 recent dated ships (full timeline below). Latest spotlight: Granite 4.1 3B (2026-04-29).

Where this lab wins

Coding: 1 tracked model with SWE-bench / HumanEval-style scores.
RAG: 1 tracked model with ruler / needle retrieval benchmarks.
Agentic: 1 tracked model with BFCL, tau-bench, and SWE-bench tool-use coverage.
Long-context: 34 tracked models with context-token or InfiniteBench-class signal.

Flagship quality / price signal

Anchor SKU: Granite 13B Chat (best sourced coding Q/$ in this portfolio).

Quality / dollar unavailable for this anchor — missing benchmark coverage and/or output token price on the cheapest ladder route (open the model detail after pricing lands).

IBM Research is an American AI research organization founded in 1945. Creating reliable and adaptable AI solutions. IBM Research ships 10 model families totaling 61 models, with the most recent release Granite 4.1 3B in 2026-04. Notable families include Granite 4.1, Granite Vision, and Granite Embedding. Use it as a stable reference for lab background, release coverage, and follow-up model pages as. View official API endpoints, benchmark performance, and coding/agent fit for every IBM Research model.

About

IBM Research, established in 1945 and based in Armonk, New York, has been an influential player in the field of artificial intelligence (AI) for decades, setting benchmarks through various historical milestones. Noteworthy achievements include the development of Arthur Samuel's self-learning checkers program in the 1950s and William Dersh's voice-operated "Shoebox" in 1962. These early innovations paved the way for the development of more sophisticated AI systems, including Deep Blue's famous 1997 chess victory over Garry Kasparov and Watson's win on Jeopardy! in 2011. These milestones are foundational to IBM's ongoing innovations in generative AI and large language models (LLMs). Currently, IBM Research focuses on creating powerful foundation models and generative AI systems that prioritize trust and transparency. This commitment is reflected in their open-source approach, designed to make AI development more accessible to enterprises. One such initiative is InstructLab, an open-source project aimed at reducing the costs associated with fine-tuning LLMs. By using Large-Scale Alignment for ChatBots (LAB) to generate high-quality synthetic data, InstructLab enhances models efficiently without necessitating complete retraining. Additionally, the Granite model series, implemented on watsonx.ai, forms the backbone for products like watsonx Assistant and watsonx Orchestrate, noted for their exceptional transparency score on Stanford's Foundation Model Transparency Index. IBM Research's innovation extends to exploring LLM routers, which can dynamically select the most cost-effective model for each query, potentially lowering inference costs by up to 85%. The collaborative efforts in these projects feature contributions from key individuals like David Cox, IBM Research's vice president for AI models, and Akash Srivastava, principal AI product advisor at Red Hat. Additionally, researchers such as Michael Muller, focused on human-centered AI, embody IBM's dedication to ethical AI development. The organization's extensive history in AI research, coupled with its current strategic initiatives, underscores its stature as a leading entity in generative AI and LLMs. IBM Research's commitment to ethical, open-source, and transparent AI systems, alongside its innovative tools like InstructLab and LLM routers, demonstrate its pivotal role in generative AI's ongoing evolution. Applications of these technologies span various domains, including drug discovery and software modernization, highlighting IBM's versatility and forward-thinking approach in the realm of AI.