Inkling Model Card - Thinking Machines Lab
Inkling Model Card
1. General information
Model name
Inkling
Legal name of the model provider
Thinking Machines Lab, Inc.
Date of release
July 15, 2026
License
Apache 2.0
Intended uses
Inkling is a general-purpose multimodal model that accepts text, image and audio inputs and generates text outputs. It is intended for use in English and other languages, and across multiple coding languages. The model is designed to be used by developers building AI-powered applications, including agentic and tool-use systems, coding assistants, chatbots, and retrieval-augmented generation systems, and is suitable for general-purpose conversational use, instruction-following, and other natural language and multimodal tasks. It is released with open weights to support research, fine-tuning and integration into third-party products by downstream developers.
Model Acceptable Use Policy
Languages
English, general multilingual support
2. Model properties
Model type
Multimodal autoregressive transformer
Architecture type
A 66-layer decoder-only transformer with a sparse Mixture-of-Experts (MoE) feed-forward backbone: each token is routed to 6 of 256 experts, plus 2 shared experts active on every token. Attention is a hybrid of local and global layers. The model is natively multimodal — images are encoded via a hierarchical patch encoder, and audio via discrete token encoding — with all modalities projected into a shared hidden space and processed jointly by the decoder.
Parameters
975B total, 41B active
Numerics support
BF16, MXFP8 and NVFP4
Input modalities
Text: UTF-8 encoded text
Image: Any pixel-based image input. For optimal performance, each image dimension should be between 40px to 4096px.
Audio: WAV format, sampled at 16kHz. For optimal performance, audio length should be within 20 mins.
Inkling supports a context window of up to 1M tokens.
Output modalities
Text: UTF-8 encoded text
3. Methods of distribution
The model is available via API access through Tinker, our service for fine-tuning. The model is also available via API access through third party inference providers.
The weights are available for download through Hugging Face.
API access to the model
For accessing our model via Tinker, you can get started by referring to the Tinker Cookbook here, and installing the tml-renderers package here.
For accessing our model via third party inference providers, you can refer to the documentation from our inference provider partners.
Running the open weights model on your own system
Required hardware
The model is distributed in two checkpoint formats.
The BF16 checkpoint requires a GPU cluster with at least 2 TB of aggregated VRAM. This can be met with either of the following configurations:
8x NVIDIA B300 GPUs
16x NVIDIA H200 GPUs
The NVFP4 checkpoint offers a quantized alternative that reduces the aggregated VRAM requirement to at least 600 GB. This checkpoint can be run as:
W4A4 on 4x NVIDIA B300 GPUs (note: W4A4 mode additionally requires SM100+ architecture)
W4A16 on 8x NVIDIA H200 GPUs
Required software
Running the model directly on GPU hardware requires an inference deployment framework–either SGLang, vLLM, TokenSpeed, Unsloth, or Hugging Face, along with all of their respective dependency libraries.
4. Training
Data types
Training data includes a broad variety of content types, including text, images, audio, video.
Data provenance
Training data for the model was drawn from publicly available sources, acquired from third-parties, or synthetically generated or augmented. Publicly available data includes content from the public internet and publicly accessible repositories.
Curation methodologies
Data curation includes cleaning, processing, and modifying datasets. These processing steps, which vary by data type, may include deduplication and filtering to remove junk or other low-quality data, or to advance safety or other objectives.
5. Evaluations
Open weights
Closed weights
Inklingeffort=0.99
Nemotron 3Ultra
Kimi K2.5
Kimi K2.6
GLM 5.2
DeepSeek V4Pro
Gemini 3.1 Pro(high)
Claude Fable 5(max)
GPT 5.6 Sol(max/xhigh)
Reasoning
HLEtext only
29.7%<br>26.6%<br>29.4%<br>35.9%<br>40.1%<br>35.9%<br>44.7%<br>53.3%<br>47.2%
HLEwith tools
46.0%<br>37.4%<br>50.2%<br>54.0%<br>54.7%<br>48.2%<br>51.4%<br>64.5%<br>55.0%
AIME 2026
97.1%<br>94.2%<br>95.8%<br>96.4%<br>99.2%<br>96.7%<br>98.3%<br>99.9%<br>99.9%
GPQA Diamond
87.2%<br>86.7%<br>87.9%<br>91.1%<br>89.5%<br>88.8%<br>94.1%<br>92.6%<br>94.1%
Agentic (coding)
SWEBench Verified*
77.6%<br>70.7%<br>76.8%<br>80.2%<br>80.0%<br>80.6%<br>80.6%<br>95.0%<br>82.2%
SWEBench ProPublic
54.3%<br>46.4%<br>50.7%<br>58.6%<br>62.1%<br>55.4%<br>54.2%<br>80.0%<br>64.6%
Terminal Bench 2.1*Best Harness
63.8%<br>56.4%<br>51.3%<br>71.3%<br>82.7%<br>64%<br>73.8%<br>84.6%<br>89.5%
Agentic (general)
GDPVal-AA v2
1238<br>1164<br>1009<br>1190<br>1514<br>1307<br>962<br>1760<br>1748
MCP Atlas
74.1%<br>44.7%<br>64.0%<br>68.1%<br>77.8%<br>73.2%<br>78.2%<br>83.3%<br>81.8%
Tau 3 Banking
23.7%<br>13.8%<br>14.2%<br>20.6%<br>26.8%<br>25.8%<br>16.5%<br>26.8%<br>33.0%
BrowseCompw/ ctx...