The Open Weights
LatestModelsLeaderboardsCompanies
Subscribe
The Open Weights

The daily record of open-source AI. New model releases, leaderboards, and what's coming next — written for people who ship.

Refreshed every 12 hours

Discover

  • Latest releases
  • New today
  • Trending models

Browse

  • All models
  • Companies
  • Categories
  • Leaderboards

About

  • About
  • Editorial policy
  • Hardware estimates
  • RSS feed
  • llms.txt
  • Newsletter

© 2026 The Open Weights. An independent publication.

PrivacyTermsSMSAggregated by Claude · curated by humans.
LatestThinkingmachines2-Small
ThinkingmachinesVision-Language

Thinking Machines ships Inkling-2-Small, an open MoE VLM

The Apache-2.0 vision-language model uses a mixture-of-experts design and lands openly on Hugging Face.

Nov 4, 2026
NotableApache 2.0
Inkling-2-Small

Thinking Machines has released Inkling-2-Small, a vision-language model published under the permissive Apache-2.0 license on Hugging Face. The model handles both image and text inputs, placing it in the growing class of multimodal systems that pair visual understanding with text generation.

Under the hood, Inkling-2-Small uses a mixture-of-experts (MoE) architecture. Rather than activating every parameter for each token, MoE models route inputs through a subset of specialized expert networks, a design that can deliver strong capability while keeping inference costs lower than a comparably sized dense model.

Why it matters

Open multimodal models remain scarcer than their text-only counterparts, and an Apache-2.0 license means developers can use, fine-tune, and deploy Inkling-2-Small commercially with minimal restrictions. That combination of open weights and permissive terms is what makes releases like this worth watching.

  • Modality: vision-language, with text generation
  • Architecture: mixture-of-experts
  • License: Apache-2.0

Thinking Machines has not published detailed specifications such as parameter counts or context length alongside the initial listing, so practical comparisons against peers will depend on independent testing. For now, the weights and model card are available directly from the Hugging Face repository.

Sources

  • thinkingmachines/Inkling-2-Small

    Hugging Face

    Visit
OlderMeituan Releases LongCat 2.5, an MIT-Licensed MoEMeituan · Text / LLM · 2 weeks agoNewerOpenBMB's Phonon Brings High-Quality TTS On-DeviceOpenBMB · Text → Speech · 2 weeks ago

Get the model

Hugging Face

Specs

Size390.1 GB
PrecisionINT8
ArchitectureInkling2ForConditionalGeneration
LicenseAPACHE-2.0
Downloads40.6K
Likes598

Can you run it?

Needs a multi-GPU server — about 232.4 GB at 4-bit.

  • INT8 (as published)

    8×H100 80 GB node

    392.7 GB
  • 4-bit

    4×H100 80 GB · 512 GB Mac Studio

    232.4 GB
Your machine
  • Mixture-of-experts: every expert must be loaded, so memory follows total parameters, not the active slice.
  • KV cache sized for 8,192 tokens of context; longer prompts need proportionally more.

Based on Hugging Face weights metadata. Assumes an 8K context and ~1 GB runtime overhead; actual needs vary. How we estimate


Modalities

Text / LLMVision-Language

The Weekly Weights

Every open release that mattered, one email a week.

0 comments

No comments yet. Be the first to weigh in.

More from Thinkingmachines

All Thinkingmachines releases →
Inkling Small
Thinkingmachines/Vision-LanguageMulti-GPU server

Thinking Machines Debuts Inkling Small, a Compact Multimodal MoE

The Apache-2.0 model brings mixture-of-experts efficiency to image, audio, and text tasks in a smaller footprint.

Jul 27, 2026
Inkling
Thinkingmachines/Any-to-AnyDatacenter

Thinking Machines Lab debuts Inkling, its first open model

The lab's inaugural open-weights release is a mixture-of-experts system that takes image and audio inputs, shipped under a permissive Apache 2.0 license.

Jul 15, 2026

More in Vision-Language

All Vision-Language →
SenseNova-Loop-9B
SenseTime/Vision-LanguageRuns on a laptop

SenseTime Releases Looped 9B Multimodal Model

The company's new SenseNova-Loop-9B pairs vision-language capabilities with a recurrent looped-transformer design.

Nov 17, 2026
Hunyuan Hy4-flash
Tencent/Vision-LanguageMulti-GPU server

Tencent releases Hunyuan Hy4-flash multimodal MoE

The lightweight, mixture-of-experts model adds vision and text instruction-following to Tencent's open Hunyuan lineup.

Nov 17, 2026
Mirage
Cactus Compute/Vision-Language

Cactus Compute's Mirage brings vision tasks on-device

The new open vision-language model targets detection, segmentation, and visual Q&A without relying on the cloud.

Nov 5, 2026