Inkling model hits Claude Code via Hugging Face

// 2h agoPRODUCT UPDATE

Inkling model hits Claude Code via Hugging Face

Thinking Machines has made its new 975-billion parameter multimodal Mixture-of-Experts model, Inkling, accessible within Claude Code. This integration is powered by Claude Code's support for Hugging Face inference providers, allowing developers to leverage the new open-weights model for their daily programming workflows.

// ANALYSIS

This integration highlights the growing trend of modular AI coding assistants that allow users to plug in their preferred open-weights models. Inkling's massive 1-million token context window and controllable thinking effort make it uniquely suited for complex, large-scale codebase analysis in Claude Code. Utilizing Hugging Face as the bridge continues to cement HF's role as the central hub for model distribution and inference routing.

// TAGS

thinking-machinesinklingclaude-codehugging-faceopen-weights

DISCOVERED

2h ago

2026-07-17

PUBLISHED

2h ago

2026-07-17

RELEVANCE

8/ 10

AUTHOR

_akhaliq

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

TUTORIAL19m ago

Microsoft "ML for Beginners" adds 50+ translations

Microsoft's popular 12-week open-source machine learning curriculum, ML for Beginners, has been updated to offer automated, always up-to-date translations into more than 50 languages, including Arabic, Hindi, and Swahili. This update aims to lower barriers to entry for aspiring machine learning practitioners globally by making the educational content accessible in their native languages.

LAUNCH53m ago

Fly.io launches Sprites, providing stateful and hardware-isolated Linux sandbox environments with fast copy-on-write checkpoint and restore capabilities.

Fly.io has introduced Sprites, which are stateful sandbox environments running in hardware-isolated AWS Firecracker microVMs designed for executing arbitrary, untrusted code or AI agents. Unlike traditional ephemeral serverless functions, Sprites retain their disk state between runs, utilizing a fast NVMe filesystem that continuously syncs to durable external storage. The platform features an ultra-fast copy-on-write checkpoint and restore system taking about 300ms, granular network egress policies using simple domain-level allowlists, and custom port forwarding for public or private service access. Sprites scale to zero and burst dynamically, meaning developers only pay for actual CPU, memory, and written storage usage.

UPDATE2h ago

Kimi Code CLI integrates Kimi K3

Kimi Code is a terminal-based AI developer CLI and agent environment from Moonshot AI that now supports their Kimi K3 flagship model. Operating locally, the tool functions as an agentic coding assistant that allows developers to run command execution, file editing, and debugging tasks without leaving the terminal, positioning itself as a competitor to terminal-native platforms like Claude Code.

Inkling model hits Claude Code via Hugging Face