Urgent.News

What's breaking now, across thousands of outlets.

AI

Artificial Analysis says Inkling 2 Mini, with 5.4B active parameters, matches Qwen3.8 27B (xhigh) on its Intelligence Index with ~5x fewer active parameters (Artificial Analysis)

Thinking Machines has released Inkling 2 Mini at 20 on the Artificial Analysis Intelligence Index, an open weights looped transformer …

Thinking Machines has unveiled Inkling 2 Mini, a 26 billion total, 5.4 billion active parameter mixture-of-experts model, achieving a score of 20 on the Artificial Analysis Intelligence Index. The model is an open weights looped transformer positioned on the Pareto frontier for Intelligence versus Active Parameters. It features three passes through its looped layers, utilizing 3.1 billion unique weights per text token, and is compact enough to run on a MacBook or DGX Spark.

Released under the Apache 2.0 license, Inkling 2 Mini supports text and image input with a 1 million token context window.

Notably, Inkling 2 Mini matches the performance of Qwen3.8 27B (xhigh) and Nanbeige4.3-33B on the Intelligence Index, with approximately 5 and 6 times fewer active parameters, respectively. It outperforms these models while using about 38% more output tokens for the same Intelligence Index score but requiring 2.8 times less compute.

Additionally, the model averages around 152,000 output tokens per Intelligence Index task, compared to 110,000 for Qwen3.8 27B and 158,000 for Nanbeige4.3-33B. On the compute proxy measure (active parameters multiplied by tokens), Inkling 2 Mini ranks on the Pareto frontier, using roughly 2.8 times less compute than Qwen3.8 27B and five times less than Nanbeige4.3-33B.

Inkling 2 Mini is approximately six times cheaper and four times faster per task than Qwen3.8 27B. It costs $0.59 and completes an Intelligence Index task in just 9.0 minutes, as opposed to $3.53 and 38.2 minutes for Qwen3.8 27B. GPT-6 Luna (max) scores 28 for $0.15 per task. Thinking Machines continues to maintain its competitive edge with its first-party pricing ($0.15/$0.50/$0.015 per 1 million input/output/cache hit tokens).

Written by urgent.news from Techmeme's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at artificialanalysis.ai →

More in AI

More from Friday 20 November →