euangelion.
Artificial intelligence

Inkling 2 Small opens on Tinker; Ripple offers downloadable coding weights

Thinking Machines launched Inkling 2 Small through Tinker with weights still planned, while Ricursive’s Apache-2.0 Ripple-65B Preview offers an immediately inspectable alternative for self-hosting experiments.

132 words · 1 min

Inkling 2 Small arrived November 5 with image input, variable reasoning effort and a mixture-of-experts architecture of 594 billion total and 17 billion active parameters.

It is available for sampling and fine-tuning on Tinker. Thinking Machines says full Apache-2.0 weights are intended for the following week after external safety testing; downloadable fine-tuned checkpoints depend on that later release. That promise should not be entered into a deployment plan as a completed weight release. [1][2]

Ricursive’s Ripple-65B Preview already has an Apache-2.0 model repository, including instruction-tuned BF16 and NVFP4 options and a separate base checkpoint. Its dense 65-billion-parameter architecture offers a smaller total model to evaluate. Ricursive’s coding comparisons mix published peer scores with its own evaluations, and its card acknowledges harness-related reproduction difficulties. Total parameter count alone cannot rank capability or serving cost. [3][4]

Sources & details 5 sources
Discuss in your voice app

Take the reporting, visual explanations and original sources into a conversation. Copy the brief, paste it into your preferred assistant, then start its voice mode.

Open ChatGPT
Paste the brief into your voice app to discuss this reporting.

Reporting sources

  1. Thinking Machines — Inkling 2 Small

    November 5 hosted release and contingent planned Apache-2.0 weight release.

  2. Thinking Machines — model card

    Model scope and limitations; same developer, no independent performance replication.

  3. Ricursive — Ripple launch

    Developer's architecture, access and self-reported development-cost claims.

  4. Ricursive — Ripple model card

    Public artifact, Apache-2.0 label, checkpoint formats and explicit harness-comparability caveat.

  5. Artificial Analysis — Inkling 2 Small evaluation

    Independent evaluator confirms proprietary hosted launch with future weights; task results show uneven coding strengths and output-token costs.

Images & figures

  • Inkling 2 Small is available through Tinker while its weight release remains planned. Ripple-65B Preview already provides Apache-2.0 weight artifacts for self-hosting experiments.

    Euangelion · Source · Usage terms