Inkling 2 Small trained in 68 hours on 12T tokens, claims edge over Muse Spark

Inkling 2 Small is a 17B-parameter sparse model (17A-594B) trained in 68 hours on 12 trillion tokens, reportedly beating Muse Spark. METR also studied AI research automation at Anthropic.

Read the full story at Latent Space ↗

Timeline · 1 report

2026-11-06 10:35 · Latent Space
[AINews] Inkling 2 Small: 17A-594B trained in 68 hours on 12T tokens, beating Muse Spark

More stories

Anthropic accelerating AI development despite CEO's slowdown calls, research group says — AP News
Anthropic's AI models improving nearly twice as fast, study finds — Reuters AI
METR audit finds Anthropic's model progress accelerating, with AI driving most of the speedup — DEV Community — Machine Learning
Artificial Analysis says Inkling 2 Small matches Gemini 3.8 Flash in its index — Techmeme
METR Finds Anthropic's Models Improving Twice as Fast, With Evidence Redacted — DailyAI
METR says Anthropic redacted much of the most informative evidence from its 100-page report — Techmeme
METR Estimates Anthropic's Model Capability Growth Has Nearly Doubled — Unite.AI
METR completes independent assessment of AI R&D acceleration at Anthropic — Alignment Forum

Get the daily brief of stories like this at 6:30 every morning →