Researchers Develop New Method to Enhance Ternary LLM Efficiency

Computer Science > Artificial Intelligence

arXiv:2609.16338 (cs)

[Submitted on 14 Sep 2026]

Title:Breaking the 1.58-bit Barrier for Ternary LLMs

Authors: Evangelos Georganas, Alexander Heinecke, Pradeep Dubey

View a PDF of the paper titled Breaking the 1.58-bit Barrier for Ternary LLMs, by Evangelos Georganas and 2 other authors

View PDF HTML (experimental)

Abstract:Ternary Large Language Models (LLM) store every weight as one of three symbols {-1,0,+1}{−1,0,+1}, so the cost of a ternary model is conventionally referenced to the information-theoretic \log_2 3 \approx 1.585log23≈1.585 bits per weight. The prevailing deployment format packs five ternary weights into one byte (five-trit packing), and due to the power-of-two group sizes used in practice this rounds up to 1.6251.625 bits per weight. This effective storage bit-width treats the three symbols {-1,0,+1}{−1,0,+1} as equiprobable. We measure the actual symbol distribution of 29 ternary LLM models and find that zeros account for up to 51.5%51.5% of all weights. Motivated by this finding, we introduce BITCOS, a simple distribution-adaptive layout comprised of a dense presence bitmap plus a compacted sign vector, and costs 2 - z2−z bits per weight element given a zero density zz in the model's weights. BITCOS stores weights more compactly than the five-trit packing in 26 of the 29 tested models, and reaches 1.4851.485 bits per weight on the sparsest of them. BITCOS is amenable to efficient unpacking on modern processors and GPUs, and we present optimized unpacking sequences for AVX-512, AVX2 and Intel Xe2 GPUs. Measured against production state-of-the-art ternary matrix-vector multiplication kernels, at the zero densities real-world ternary models exhibit, the realized gain with our proposed layout is up to 1.28\times1.28×. Finally, we illustrate end-to-end LLM inference results on 5 different platforms (client and server CPUs, integrated and discrete Xe2 GPUs) where decode throughput improves by up to 1.18\times1.18× on CPUs and 1.27\times1.27× on GPUs.

Subjects:	Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
Cite as:	arXiv:2609.16338 [cs.AI]
	(or arXiv:2609.16338v1 [cs.AI] for this version)
	https://doi.org/10.48550/arXiv.2609.16338Focus to learn morearXiv-issued DOI via DataCite (pending registration)

Submission history

From: Evangelos Georganas [ view email ]

[v1] Mon, 14 Sep 2026 20:54:24 UTC (144 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled Breaking the 1.58-bit Barrier for Ternary LLMs, by Evangelos Georganas and 2 other authors

• View PDF
• HTML (experimental)
• TeX Source

view license

Current browse context:

cs.AI

new | recent | 2026-09

Change to browse by:

cs

cs.LG

References & Citations

• NASA ADS
• Google Scholar
• Semantic Scholar

export BibTeX citation

Bookmark

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer ( What is the Explorer?)

Connected Papers Toggle

Connected Papers ( What is Connected Papers?)

Litmaps Toggle

Litmaps ( What is Litmaps?)

scite.ai Toggle

scite Smart Citations ( What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv ( What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers ( What is CatalyzeX?)

DagsHub Toggle

DagsHub ( What is DagsHub?)

GotitPub Toggle

Gotit.pub ( What is GotitPub?)

Huggingface Toggle

Hugging Face ( What is Huggingface?)

ScienceCast Toggle

ScienceCast ( What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate ( What is Replicate?)

Spaces Toggle

Hugging Face Spaces ( What is Spaces?)

Spaces Toggle

TXYZ.AI ( What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower ( What are Influence Flowers?)

Core recommender toggle

CORE Recommender ( What is CORE?)

• Author
• Venue
• Institution
• Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax ( What is MathJax?)