1. X
  2. Nebius Token Factory
Log inSign up
Nebius Token Factory
Nebius
951 posts
user avatar
Nebius Token Factory
Nebius
@nebiustf
Own your intelligence. Precision-built inference for AI workloads at scale. Powered by @nebiusai | Discord: discord.com/invite/WJ2DUQR…
The Netherlands
tokenfactory.nebius.com/?utm_medium=X&…
Joined November 2024
588
Following
11K
Followers
RepliesRepliesArticlesArticlesMediaMedia

Log in or sign up for X

See what’s happening and join the conversation

Continue with phone
or
Log in with username or email
Terms·Privacy·Cookies·Accessibility·Ads Info·© 2026 X Corp.
  • Pinned
    user avatar
    Nebius Token Factory
    Nebius
    @nebiustf
    Jul 27
    Kimi K3 is now available on Token Factory. We’re excited to announce that Nebius Token Factory is an official Day 0 partner for @Kimi_Moonshot's Kimi K3. Kimi K3 is the first open-weight model to reach frontier-level performance, a major step forward for open models. It is
    380K
  • user avatar
    Nebius Token Factory
    Nebius
    @nebiustf
    Jul 24
    👀 👀
    user avatar
    dylan ツ
    Nebius
    @demian_ai
    Jul 24
    Kimi K3 coming to @nebiustf in a few days
    GIF
    62K
  • user avatar
    Nebius Token Factory
    Nebius
    @nebiustf
    Jul 22
    You can now run full parameter supervised fine-tuning across three @GoogleDeepMind models: • Gemma 4 31B • Gemma 4 E4B • Gemma 4 E2B Bring your own data and adapt Gemma 4 to your domain, use case, and workflows, all through Token Factory. Start fine-tuning:
    user avatar
    Nebius Token Factory
    Nebius
    @nebiustf
    Jul 20
    DeepSeek V4 Flash is now available for supervised fine-tuning on Nebius Token Factory. Bring your own training data, adapt the model to your workflow, and run full-parameter SFT through a managed Post-training platform. Train directly from your existing object storage. No data
    6.3K
  • user avatar
    Nebius Token Factory
    Nebius
    @nebiustf
    Jul 21
    Speculative decoding works when speed and acceptance move together. SlimSpec cuts LM-head cost 4–5× without cutting the vocabulary, delivering up to 8–9% higher end-to-end speedup vs baselines. More useful tokens/second, better production throughput. See the results:
    5.5K
  • user avatar
    Nebius Token Factory
    Nebius
    @nebiustf
    Jul 20
    Come and battle datamon in the Charlie and the Token Factory 🎮
    user avatar
    dylan ツ
    Nebius
    @demian_ai
    Jul 20
    Kimi K3 ONE SHOTTED an entire playable game for me: 2 sentence prompt in, complete browser RPG out. how Kimi pulled it off is WILD. The 1st prompt made the game, then the next 4 made it mine. This was my entire opening brief: “Create a web version of Pokémon, but in a world of
    00:00
    16K