hermes-ai.net

Read docs →
HermesHermes Agent Docs
News
Back to News
repoOpen SourcePost Training

Bespoke Labs' Nimble Beats a 27B Model at Classification Using Just 9B

Bespoke Labs open-sources Nimble, a 9B model that makes typed decisions without generating tokens, using contrastive pairs for training.

Bespoke Labs' Nimble Beats a 27B Model at Classification Using Just 9B
Source
Bespoke Labs
Published
Author
AlphaSignal Newsroom
Read
1 min read

Bespoke Labs open-sources Nimble, a 9B model that makes typed decisions without generating tokens, using contrastive pairs for training.

Reporting is indexed from AlphaSignal. Rights remain with the original publisher and cited sources.

Read original report

Next reads

RadixArk's Miles v0.1 Cuts Trillion-Parameter RL Sync Time by 86%
AlphaSignal · paper

RadixArk's Miles v0.1 Cuts Trillion-Parameter RL Sync Time by 86%

Jared Palmer's Open-Source Kev Answers Typed Questions in one Pass
AlphaSignal · repo

Jared Palmer's Open-Source Kev Answers Typed Questions in one Pass

Open-Source Laya Beats Closed Jev API at 7x Faster Decision Speed
AlphaSignal · repo

Open-Source Laya Beats Closed Jev API at 7x Faster Decision Speed