Inkling

An open-weights multimodal model by Thinking Machines Lab — 975B total parameters, 41B active, built for customization, agentic coding, and controllable thinking.

975B Total 41B Active 1M Context Apache 2.0 MoE
45T
Training Tokens
1M
Context Window
256
Routed Experts
Text · Image · Audio
Modalities

Try Inkling

Chat with Inkling through the Hugging Face Inference API. Powered by Thinking Machines Lab's open-weights model.

Demo by akhaliq — uses the HF Inference Providers router.

Capabilities

Inkling is a broad generalist foundation model trained for real-world customization.

🧠

Controllable Thinking

Adjust effort from 0.2 to 0.99 to balance performance vs. token cost. Reaches the same score as Nemotron 3 Ultra on Terminal Bench 2.1 at ~1/3 the tokens.

💻

Agentic Coding

Scores 77.6% on SWE-Bench Verified and 63.8% on Terminal Bench 2.1. Builds web apps, games, and PDFs through tool use and long refinement loops.

👁️

Vision & Audio

Native multimodal reasoning over images, charts, diagrams, and audio. Encoder-free architecture using dMel spectrograms and hMLP patches.

🎯

Calibrated Epistemics

Trained with proper scoring rules for calibrated confidence. Strong instruction following via rubric + claims graders with agentic web search.

🛡️

Safety-First

Among the strongest open-weights safeguards on FORTRESS (78% adversarial) and StrongREJECT (98.6%). Commissioned external safety testers.

⚙️

Fine-Tunable

Available on Tinker for fine-tuning. Inkling can even self-fine-tune — it wrote and ran its own lipogram fine-tuning job and switched weights.

Benchmarks

Scored at effort=0.99 across reasoning, coding, agentic, factuality, and multimodal tasks.

Reasoning
BenchmarkInklingNemotron 3 UltraKimi K2.6GLM 5.2
HLE (text only)29.7%26.6%35.9%40.1%
HLE (with tools)46.0%37.4%54.0%54.7%
AIME 202697.1%94.2%96.4%99.2%
GPQA Diamond87.2%86.7%91.1%89.5%
Agentic (Coding)
BenchmarkInklingNemotron 3 UltraKimi K2.6DeepSeek V4 Pro
SWE-Bench Verified77.6%70.7%80.2%80.6%
Terminal Bench 2.163.8%56.4%71.3%64%
SWE-Bench Pro Public54.3%46.4%58.6%55.4%
Multimodal
BenchmarkInklingQwen3-OmniKimi K2.6Gemini 3.1 Pro
MMMU Pro (vision)73.5%60.0%75.0%82.0%
MMAU (audio)77.2%77.5%82.5%
VoiceBench91.4%88.8%94.3%
Safety
BenchmarkInklingNemotron 3 UltraKimi K2.6DeepSeek V4 Pro
FORTRESS (Adversarial)78.0%77.6%65.6%36.0%
StrongREJECT98.6%98.7%99.8%98.6%

Scores from the official Inkling announcement. Closed-weight models (Claude Fable 5, GPT 5.6 Sol, Gemini 3.1 Pro) omitted for brevity. See the full post for complete comparisons.

Architecture

Inkling is a Mixture-of-Experts Transformer designed for efficient long-context performance.

975B Total · 41B Active

256 routed experts + 2 shared experts per MoE layer, with 6 experts active per token. Sigmoid-based router with auxiliary-loss-free load balancing.

Hybrid Attention

Sliding-window and global attention layers interleaved at a 5:1 ratio with 8 KV heads. Relative positional embeddings instead of RoPE for better long-context extrapolation.

45 Trillion Tokens

Pre-trained on text, images, audio, and video. Muon optimizer for large matrix weights, Adam for others. RL scaled to over 30M rollouts for post-training.

Resources

Get started with Inkling — fine-tune, deploy, and build with it.