Thinking Machines Lab releases Inkling: 975B-parameter open-weights multimodal model under Apache-2.0
Thinking Machines Lab released Inkling, a Mixture-of-Experts transformer with 975B total parameters and 41B active parameters, trained on 45 trillion tokens of text, images, audio and video. The Apache-2.0 licensed model is designed as a base for fine-tuning rather than a frontier model.
Inkling — Quick Specs
Thinking Machines Lab releases Inkling: 975B-parameter open-weights multimodal model under Apache-2.0
Mira Murati's Thinking Machines Lab released Inkling, a Mixture-of-Experts transformer with 975B total parameters and 41B active parameters. The multimodal model was trained on 45 trillion tokens spanning text, images, audio, and video, and is released under the Apache-2.0 license.
According to Thinking Machines Lab, Inkling is not positioned as a frontier model. Instead, the company describes it as "a good open-weights base for customization" featuring "multimodal capabilities, efficient thinking, and availability on Tinker for fine-tuning." The model is specifically designed for use with the company's Tinker training platform.
The lab also announced Inkling-Small, a 276B-parameter model with 12B active parameters, though weights will only be released "once that work is complete" after additional testing.
Training data documentation minimal
The model card and training data documentation provide limited detail about the training process. The training data documentation states only that datasets include "content that is in the public domain as well as content that may be subject to intellectual property protection" and were "obtained from the open internet and publicly accessible data repositories." The company notes that "certain datasets were also obtained from third parties" but provides no additional specifics.
Model capabilities demonstrated
The model supports both text generation and multimodal understanding through the Thinking Machines API. When prompted to generate an SVG of a pelican riding a bicycle, the model produced vector graphics output. However, when asked to describe the resulting image, the model identified the bird as "a stork or seagull" rather than recognizing it as a pelican.
Pricing for API access has not been disclosed. The model weights are available for download, though specific benchmark scores have not been published.
What this means
Inkling adds another Apache-2.0 licensed option to the US open-weights ecosystem, joining models like NVIDIA Nemotron and Gemma 4. The 41B active parameter count from a 975B total suggests an efficiency-focused architecture, though the lack of benchmark data makes direct performance comparisons difficult. The minimal training data documentation continues a trend of limited transparency from some AI labs, particularly regarding potential use of copyrighted content. As a fine-tuning base rather than a general-purpose model, Inkling's real value will depend on whether developers adopt the Tinker platform for customization.
Related Articles
Mistral AI Releases Shieldstral-1.0-3B, a 3B-Parameter Policy-Adaptive Safety Classifier
Mistral AI has released Shieldstral-1.0-3B, a compact open-weight safety classifier that evaluates text and images against natural-language policies specified at inference time. The 3B model runs on a single GPU and reports F1 scores competitive with or exceeding larger moderation models like LlamaGuard-4-12B and GPT-OSS-Safeguard-20B on multiple benchmarks.
Mistral Releases Shieldstral, a 3B Open-Weights Safety Classifier That Matches Models 7x Its Size
Mistral has released Shieldstral, a 3B open-weights safety classifier that reframes content moderation as a policy-adaptive question-answering task. The model claims to match or outperform guard models up to 7x its size on text safety and multimodal benchmarks, and runs on a single 16GB GPU.
LG AI Research Releases K-EXAONE 2.0, a 750B-Parameter Open-Weight MoE Model with 262K Context
LG AI Research has released K-EXAONE 2.0, a 750-billion-parameter mixture-of-experts language model with 37B active parameters, a 262,144-token context window, and support for 10 languages. The model is open-weighted under Apache 2.0 and claims competitive results against Qwen3.5, GLM-5.1, and DeepSeek-V4 Pro on reasoning, coding, and long-context benchmarks.
Alibaba Unveils Qwen3.8-Max, a 2.4T-Parameter Open-Weight Model for Coding and Agentic Work
Alibaba announced Qwen3.8-Max, a 2.4T-parameter flagship model targeting coding and long-horizon agentic work, with open weights promised for next week alongside Qwen3.8-27B. The model posted strong third-party benchmark results, ranking #4 in Frontend Code Arena and matching Claude Opus 4.7 on the Vals Index at roughly 2.3x lower cost.
Comments
Loading...