Chat with Inkling Small.

Thinking Machines Lab's smaller model. Wakes 12B of its 276B per word, so it answers quickly, and it reads pictures.

Start chatting with Inkling Small now — free, no account

At a glance

  • Maker: Thinking Machines
  • Released: 2026
  • Parameters: 276B
  • Active per token: 12B
  • Context window: 524k
  • Size: Very large
  • Memory: Handles very long documents
  • Good at: Quick multimodal answers
  • Supports: tools, images, zero data retention
  • License: Apache 2.0
  • Upstream API price: $1.20 per million tokens

Announced in Introducing Inkling-Small.

What is Inkling Small?

Inkling Small is Thinking Machines' smaller model — the quick sibling of the lab's flagship Inkling, and the entry point to a line built by one of the more interesting young labs in open AI. Thinking Machines Lab was founded by former OpenAI researchers with a publish-everything ethos, and the Inkling pair shows the philosophy in miniature: capable open weights, documented behaviour, no mystique.

The numbers describe a classic modern mid-fast model: 276 billion parameters in total with only about 12 billion awake per token — a sparsity ratio that keeps answers quick and the price at $1.20 per million tokens. The context window is a distinctive 524k tokens, roughly 390,000 words, sitting between the million-token giants and the quarter-million mid-field. It reads pictures as well as text, supports tools, and is covered by zero-data-retention terms.

Its stated speciality is quick multimodal answers — the fast lane of a lab that also builds a flagship. If you like the Inkling temperament but find the full model's $4.05 hard to justify for daily work, Small is the one you actually keep loaded.

What can you do with Inkling Small?

A quick multimodal mid-fast model finds plenty of work:

  • Fast answers about pictures — screenshots, charts, photos — in conversational flow.
  • Mid-length document work inside the 524k-token window, roughly 390,000 words.
  • Coding help where speed matters more than ultimate depth.
  • Agent workflows with tools at a moderate per-turn cost.
  • A daily default from a lab whose flagship you keep for hard problems.
  • Comparing quick multimodal models — Small against the Flash tier of other families.

How does Inkling Small compare?

Against its big sibling Inkling ($4.05), it is the speed-versus-depth trade at a third of the price: 12B awake per token versus a full 975B model's deliberation, 524k versus one million tokens of context, pictures versus pictures-and-sound. Small is the default; Inkling is the escalation.

Across makers, the same-price comparison is MiniMax M3 ($1.20) — bigger, longer-context, video-capable but currently paused — while Gemma 4 26B ($0.40) undercuts it with a smaller window. Inkling Small's pitch is balance: quick, sighted, decently long-memoried, from a lab you can read along with.

Is Inkling Small right for you?

Inkling Small is the pick when you want the Inkling temperament at a daily-driver price: quick multimodal answers, pictures read natively, tools supported, and a 524k-token context window that is genuinely its own niche — larger than the quarter-million mid-field, smaller than the million-token giants, and in practice the right size for most real documents with room to spare. At $1.20 per million tokens it is priced like a mid-field model while behaving like a small flagship.

The escalation path is the point of the pair. Small at $1.20 for the day's work, the full Inkling at $4.05 for the hard problems, the same lab's reasoning character through both — many two-model setups are really one philosophy with a fast lane and a slow lane, and the Inkling pair is that idea done cleanly. When Small hesitates on something that matters, you already know where the answer goes.

Against the field, Small holds its own on breadth — sight plus tools plus a distinctive context size — and concedes on price to the specialists: Gemma 4 26B undercuts it for quick multimodal work, and MiniMax M3 matches the price with more reach when it is selectable. What Small offers instead is balance, and a lab whose research you can read — which counts for more than people expect when something behaves oddly in production and you want to know why.

Frequently asked questions

How does Inkling Small compare to Inkling?

Small is the quick sibling: 276B parameters with 12B awake per token, a 524k context, picture input, at $1.20 per million tokens. The full Inkling is the flagship: 975B parameters, a million-token window, sound input too, at $4.05. Small for daily work, Inkling for the hardest problems. The 3.4x price multiple sounds stark until you price a week of daily work: at Small’s $1.20, most people never spend enough to justify the flagship.

What is the context window of Inkling Small?

524k tokens — roughly 390,000 words. A distinctive middle ground: more than the quarter-million mid-field, less than the million-token giants. That window is the model’s quiet weapon — large enough for serious documents, small enough that the model stays quick rather than wading through padding. It is the size of window that makes long agent loops practical without paying for a million tokens you never fill.

Does Inkling Small support image input?

Yes — it reads pictures as well as text. Audio and video input belong to the full Inkling, not to Small. Chart reading, screenshot triage and photo questions are the everyday cases, and it handles them with the family’s characteristic care.

Who makes Inkling Small?

Thinking Machines Lab, a research company founded by former OpenAI researchers that publishes its work and releases open weights. Inkling Small has its own announcement post on the lab's site. The lab publishes its evaluations alongside its models, which is rarer than it should be and genuinely useful when you are deciding what to trust.

Why is it called Inkling Small?

Because it is the smaller sibling of Inkling, the lab's larger model — though small is relative: 276 billion parameters in total, with a sparse 12B awake per token for speed.

Is Inkling Small free to chat with on UsingOpen?

Yes. Pick it in the model picker and start chatting — no account, no payment, with history saved only on your device. It is an easy model to underestimate on paper and a hard one to dislike in practice — one session usually settles it.

Related models on UsingOpen

Models most often compared with Inkling Small on UsingOpen:

  • Inkling by Thinking Machines: the larger Thinking Machines model — 975B parameters that read text, pictures and sound.
  • MiniMax M3 by MiniMax: the multimodal agent — 427B parameters, a million-token window and native video.
  • Gemma 4 26B by Google: the mixture-of-experts Gemma — 25B total, 3.8B awake, quick on a budget.
  • GLM 5.3 Flash by Z.ai: the fast default — 18B of 320B parameters awake per token, a million-token window and picture input.

Head-to-head comparisons

See how Inkling Small lines up with the models people most often weigh it against:

Model availability changes as upstream capacity comes and goes — the picker on the chat page shows what is ready right now.

Chat with Inkling Small free on UsingOpen — no account, no sign-up

← All open-weight models · Compare models · Free AI chat · FAQ