PulseAugur
EN
LIVE 14:56:05

FreeToken challenges Ollama and llama.cpp with new MoE serving engine

By PulseAugur Editorial · [1 sources] ·

FreeToken, a new Mixture of Experts (MoE) model serving engine, has been released and compared against established tools like Ollama and llama.cpp. Unlike Ollama and llama.cpp, which split model weights between GPU and CPU at load time, FreeToken employs a dynamic approach. It keeps all experts in CPU RAM and uses the GPU's spare memory as an LRU cache for recently used experts, intelligently deciding whether to pull weights over PCIe or compute them on the CPU based on real-time bandwidth measurements. AI

IMPACT Introduces a novel serving engine for MoE models that may offer performance improvements on resource-constrained hardware.

RANK_REASON New software tool release for model serving.

Read on Towards AI →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

FreeToken challenges Ollama and llama.cpp with new MoE serving engine

How we ranked this

Signal score
51 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
New software tool release for model serving.
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
infra
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. Towards AI TIER_1 English(EN) · Abhinandan Malhotra ·

    FreeToken Just Shipped. Is It Better Than Ollama and llama.cpp?

    <h4>Testing the radical new MoE serving engine against the reigning champions on a 6GB laptop</h4><p>FreeToken shipped recently. As a new serving engine for Mixture of Experts (MoE) models, its headline claim of running a 753B parameter model on a single GPU garnered a lot of att…

RELATED ENTITIES

RELATED TOPICS