Newsverum.
SearchK

Search stories

Find the story. Follow the source.

Explore
Explore the data1840Agriculture232Business287Climate46Culture12Economy264Education60Energy427Environment44Finance62Fiscal Policy46Government51Health96Infrastructure88Politics18Science25Security33Society14Sports29Technology310
EN
العربيةDeutschEnglishEspañolFrançaisPortuguêsTürkçe
Sections
Technology / WORLD

Liquid AI releases 24-billion-parameter LFM2.5 model

The company says two billion parameters are active per token and reports device-specific results.

Newsverum · Source brief1 min read

Event / data period: 2026-10-13

Published: · Updated:

Reading sizeSources ↓Share story ↗
Conceptual illustration of technology, processors and connected information.Illustration
Symbolic illustration

Liquid AI released LFM2.5-24B-A2B on October 13 as an open-weight reasoning model for local agents. The mixture-of-experts model has 24 billion total parameters and activates two billion per token. Liquid AI says its context window grew from 32,768 to 131,072 tokens, vocabulary from 65,000 to 128,000, and pretraining from 17 trillion to 33 trillion tokens versus its predecessor.

The company reports 184 generated tokens per second on an Apple M5 Max and more than 21,000 output tokens per second on one NVIDIA B300. The latter is a high-concurrency sustained-load result, not a single-user rate; both figures come from Liquid AI’s benchmarking setup. Its benchmark table shows mixed results, with the model ahead or tied on five of eight tool-use tests but behind larger models on several knowledge and coding tests.

Data

Timeline

Announcement summary

The company says two billion parameters are active per token and reports device-specific results.

  1. · Occurred

    Liquid AI releases 24-billion-parameter LFM2.5 model

    The company says two billion parameters are active per token and reports device-specific results.

Methodology

This event record summarizes the source announcement and adds no independent verification or generalization beyond it.

Updated:

DownloadCSV ↓JSON ↓
Structured coverage

Subjects and places

Subject proposals

Geographic references

Only source-provided geographic references are shown. Opening an external map is optional. Optional story maps use OpenStreetMap. Opening a map sends your IP address and this site’s address to the map provider; location data remains readable as text without loading the map.

  • United States · Place role: Coverage · Place type: Country or area · Location precision: Regional
    Evidence
    • Explicit preserved editorial country/region; no city or national centroid inferredwww.liquid.ai ↗
Taxonomy version: IPTC Media Topics 2026-Q2

Continue reading

Technology

Microsoft reports completion of MAI-Base-2 pre-training

Microsoft AI said on October 15 that it completed pre-training MAI-Base-2, a mixture-of-experts model with 2.8 trillion total and 80 billion active parameters. The company says training used 16,000 GB200 GPUs over 38 days in August and September; the model is still undergoing post-training.

Microsoft AI1 min read
Data included
Technology

AWS reports MoE kernel optimization results on SageMaker

AWS detailed GPU kernel optimizations for mixture-of-experts models on SageMaker AI on October 13. The company described swapping matrix operands, routing FP8 operations to the Triton backend and removing unnecessary just-in-time recompilations.

AWS Artificial Intelligence Blog1 min read
Data included
Technology

Meta releases Muse Spark Open model weights

Meta released weights for Muse Spark Open, a multimodal model it describes as having 816 billion total parameters and about 42 billion active per token.

Meta AI Research1 min read
Data included