International
The Morning Dev

Only the news worth knowing

‹
 

Research3 sources

Inkling 2 Small: Thinking Machines' 594B Model That Scores Like a Frontier System

DEV · By NeurAITool ·

The Brief · written from 3 sources

  • An analysis by Vals AI found Inkling 2 Small swore 316 times across 2,655 benchmark runs, according to The Decoder.
  • The profanity typically appears late in long runs that fail anyway.
  • Thinking Machines Lab released Inkling 2 Small on November 5, 2026, with 594B total parameters and 17B active per token.

Why it matters

According to the analysis, the profanity does not work as a warning signal for failed runs, so teams should not rely on it to flag bad reasoning output.

Who’s affected

  • Developers evaluating Inkling 2 Small for pipelines
  • Teams that monitor model reasoning traces

What to do

  1. Confirm whether the full weights are available before building a pipeline on Inkling 2 Small

Mira Murati's Thinking Machines Lab has shipped its second model: 594 billion parameters, a score of 31 on the Artificial Analysis Intelligence Index — more than double the first Inkling Small — and a promise that the full weights could be free to download under Apache 2.0 next…

Read the full story at dev.to ↗

Timeline

  1. Thinking Machines Lab releases Inkling 2 Small

  2. Vals AI analysis of profanity in benchmark runs is reported

  3. The Decoder

    Thinking Machines' Inkling 2 Small drops 316 F-bombs across benchmark runs

  4. DEV Community: ai

    Inkling 2 Small: Thinking Machines' 594B Model That Scores Like a Frontier System

  5. DEV Community: ai

    Inkling 2 Small: A Developer's Checklist Before You Bet a Pipeline on a Model With No Weights Yet

Full coverage · 3 sources

How do you feel about this?

0 Comments

    Inkling 2 Small: Thinking Machines' 594B Model That Scores Like a Frontier System – The Morning Dev