thinkingmachines/Inkling · Hugging Face

Inkling is a general-purpose multimodal model released by Thinking Machines Lab with open weights under the Apache 2.0 license. It is designed to handle text, image, and audio inputs and generate text outputs, making it suitable for conversational AI, instruction-following, coding assistants, agentic systems, retrieval-augmented generation, and other multimodal applications. The model supports English and has broader multilingual capabilities, as well as use across multiple coding languages. Technically, Inkling is a 66-layer decoder-only autoregressive transformer with a sparse Mixture-of-Experts architecture. Each token is routed to 6 of 256 experts, with 2 shared experts always active. It uses hybrid local and global attention, and processes images and video through a hierarchical patch encoder while audio is encoded as discrete tokens in a shared hidden space. The model has 975 billion total parameters, with 41 billion active at inference time, and supports BF16 and NVFP4 numerics. Inkling accepts UTF-8 text, pixel-based images, and 16 kHz WAV audio, with optimal image sizes between 40 and 4096 pixels per dimension and audio under 20 minutes. Training data came from public internet sources, public repositories, third-party data, and synthetic or augmented data, with cleaning, deduplication, and filtering applied for quality and safety. Evaluation results show strong performance on reasoning, math, knowledge, and coding benchmarks, including HLE, AIME 2026, GPQA Diamond, and SWE-bench Verified. The model is available through Tinker, local deployment frameworks such as SGLang and vLLM, and third-party inference providers.

Build this article’s timeline

Extract dated events, key actors, and open questions using the same timeline analysis available in Signal.

Log in to build

Timeline generation is available to signed-in community members.

Five Ws 🤖 5 AI prefills · all filled
Who
Hugging Face, Thinking Machines Lab, and Tinker
🤖 AI prefill
What
Inkling is a general-purpose multimodal model released by Thinking Machines Lab with open weights under the Apache 2.0 license.
🤖 AI prefill
When
2026 and Jul 14, 2026
🤖 AI prefill
Where
English, WAV, and UTF-8
🤖 AI prefill
Why
Centers on multimodal model, Mixture-of-Experts, and decoder-only transformer
🤖 AI prefill

Login to answer and earn XP for each gap you fill.

Model Overview Inkling, multimodal, open weights, text, image Deployment Access Tinker, SGLang, vLLM, TokenSpeed, Unsloth Architecture Specs transformer, Mixture-of-Experts, decoder-only, 66-layer, parameters Training Data training data, public internet, repositories, synthetic, augmentation Benchmark Results evaluations, HLE, AIME, GPQA, SWEBench
multimodal model99% Mixture-of-Experts97% decoder-only transformer95% autoregressive transformer93% open weights91% instruction-following89% retrieval-augmented generation88% agentic and tool-use systems87% hierarchical patch encoder86% shared hidden space82%
1
Manipulation Score: Low How much this article uses emotional framing vs. straight reporting
Loaded Language 025%
Absolutist 215%
Threat/Panic 025%
Us vs Them 015%
Engagement Bait 220%
Manipulation score look right?

Factual Claims (10)

low technology
Inkling is a general-purpose multimodal model that accepts text, image, and audio inputs and generates text outputs.
low technology
The model is intended for use in English and other languages, including multiple coding languages.
low technology
Inkling is designed for developers building AI-powered applications such as agentic systems, coding assistants, chatbots, and RAG systems.
low technology
The model is released with open weights for research, fine-tuning, and integration into third-party products.
low technology
Inkling is a multimodal autoregressive transformer.
medium technology
It uses a 66-layer decoder-only transformer with a sparse Mixture-of-Experts backbone.
medium technology
Each token is routed to 6 of 256 experts, plus 2 shared experts active on every token.
medium technology
The model has 975B total parameters and 41B active parameters.
low technology
Training data includes text, images, audio, and video from public, third-party, and synthetic sources.
medium technology
Inkling's reported benchmark results include 97.1% on AIME 2026 and 87.2% on GPQA Diamond.

Comments

Login to view and post comments.

🏠Portal 📰Links ❓Q&A 📅Events 💼Jobs