HOT RADAR
llmHEAT 88.6°

AI CURATED HOT STORY · 11/20/2026

Thinking Machines Open-Sources 26B Recurrent Model Inkling 2 Mini

6 reports archived6 independent sourcesupdated 11/21/2026, 07:24:30
Synthesis & Latest Updates
6 Sources Cross-Validated

Thinking Machines Lab, in collaboration with Hugging Face, has open-sourced Inkling 2 Mini, a multimodal mixture-of-experts model with 26B total parameters but only 3-5B activated per token. It scores 20 on the Artificial Analysis Intelligence Index and supports a 1M token context with text and image inputs, running locally on a MacBook or DGX Spark.

The model's key innovation is a recurrent Transformer architecture: the middle 12 layers share weights and loop three times, each iteration maintaining its own KV cache and reading gated messages from the previous round, thereby boosting effective capacity without increasing parameter count. vLLM provides Day-0 support on NVIDIA Blackwell, Hopper, and DGX Spark, with NVFP4/MXFP8 mixed-precision weights and a built-in DSpark speculative decoder for further acceleration. According to Hacker News, it achieves 48.2% on DeepSWE v1.1 and 46.5% on Vals Index v2.2, claiming leadership among all open-weight models under 100B parameters.

The release marks the arrival of efficient recurrent architectures in the open-source community, offering new possibilities for local deployment and long-context applications, and is poised to drive edge AI and privacy-sensitive scenarios while pressuring conventional Transformer models of similar scale.

LATEST/Following its release, Inkling 2 Mini quickly gained Day-0 support from vLLM, running efficiently on NVIDIA Blackwell, Hopper, and DGX Spark. Community attention is focused on the actual inference efficiency gains from the recurrent layer design and on-device deployment performance under the Apache 2.0 license.

Hourly heat curve

35 fully observed hours
Now
88.6
Peak
112.008:00
24h change
+179%

The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.

Coverage timeline

Total 6 reports · Latest first
  1. Artificial Analysis (@ArtificialAnlys)T1·78 pts
    • 26B total and 5.4B active parameters, with three passes through looped layers and only 3.1B unique weights per token
    • Scores 20 on the Artificial Analysis Intelligence Index, on the Pareto frontier for intelligence vs. active parameters
    • Apache 2.0 licensed with text and image input plus a 1M-token context, runnable locally on a MacBook or DGX Spark