Thinking Machines Launches Inkling
ArtificialAnlys · x · 2026-07-16
Thinking Machines has launched Inkling, its first production-grade language model and a new leader among US open-weight models.
Key Details
- Scale: 975B total parameters, 41B active parameters
- Modalities: Supports text, image, and audio inputs; outputs text
- Context: 256K on the Tinker API; 1M for the open-weight version
- Pricing: At 64K context, $1.87 per 1M input tokens and $4.68 per 1M output tokens; higher at 256K
Benchmark Performance
- Scored 41 on the Artificial Analysis Intelligence Index, surpassing the previous best US open-weight model, Nemotron 3 Ultra (38)
- Excels in agentic tasks, leading some competitors on GDPval-AA v2 and 𝜏³-Banking
- Highly token-efficient, with lower average output token counts compared to GLM-5.2, Kimi K2.6, and DeepSeek v4 Pro
Technical Implementation
- Images/videos are encoded via a hierarchical patch encoder
- Audio is encoded using discrete tokens
- All modalities are projected into a shared hidden space and jointly processed by a decoder
More from Models
- Mocked benchmark shows DeepSeek's document chart grounding badly misaligned — burkov · 2026-10-18
- Dev complains Claude's habit of saying "not done" is his biggest pain point — jasonkneen · 2026-10-18
- Open-weight TYGER speech decision model makes 80ms turn-taking calls for full-duplex voice agents — solyarisoftware · 2026-10-18
- Paying Google AI subscriber blocked: custom-domain Workspace accounts can't access AI Studio — ChrisRackauckas · 2026-10-18
- Running Ling 3.1 Flash (560B MoE, 25B Active) Across 4 DGX Sparks via llama.cpp RPC — BerryBlush_Rogers691 · 2026-10-18
- Users press OpenAI on long-delayed Adult Mode a year after it was promised — ___Patrice___ · 2026-10-18