model comparison

vs

Inkling vs Inkling Small

two Thinking Machines models, compared on listed price, real workload cost, context and capabilities.

the short answer

  • on SWEBench Verified, as each maker reports it, Inkling scores 77.6% and Inkling Small 80.2%.
  • Inkling Small lists lower input and output prices.

side-by-side specs

specInklingInkling Small
providerThinking MachinesThinking Machines
input price$1/M$0.5/M
output price$4.05/M$1.2/M
cached input$0.17/M—
blended price (3:1)$1.762/M$0.675/M
context window256k tokens256k tokens
inputtext, imagestext, images
structured outputyesyes
weightsopen (Apache 2.0)open (Apache 2.0)
releasedJuly 15, 2026July 30, 2026
SWEBench Verified (maker-reported)77.6%80.2%
Terminal Bench 2.1 (Best Harness) (maker-reported)63.8%64.7%

same work, both models

workloadInklingInkling Small
one chat reply3,000 tokens in, 500 outInkling costs 2.4× as much as Inkling Small$0.005$0.0021
summarise a 100-page report50,000 tokens in, 1,500 out, one requestInkling costs 2.1× as much as Inkling Small$0.06$0.03
1,000 classification calls500 tokens in, 20 out, per call$0.58$0.27
coding-agent session2M tokens in across ~60 turns, 60,000 out, no cache$2.24$1.07

at listed rates. token counts are typical sizes, not measurements.

when to pick which

Inkling

Inkling is Thinking Machines' open-weights model and the first in its family: a 975B-parameter mixture-of-experts (41B active) that reasons over text, images and audio with controllable thinking effort. Inkling Small is its smaller, cheaper sibling.

pick it if

  • you want Apache 2.0 open weights you can later fine-tune on Tinker
  • you send images with text and want to tune thinking effort
  • you run coding or agent harnesses, the work it was trained for

skip it if

  • cheaper Inkling Small, at half the input price, covers your workload
  • you need the strongest model; Thinking Machines says Inkling is not it

Inkling Small

Inkling Small is Thinking Machines' 276B-total, 12B-active mixture-of-experts open-weights model, the smaller sibling of Inkling. the lab reports performance comparable to Inkling at a quarter of its size; on octomind it costs half Inkling's input rate and under a third of its output rate.

pick it if

  • want the Inkling family's lower-priced tier with image input
  • need adjustable thinking effort, from minimal to xhigh
  • want Apache 2.0 open weights you can fine-tune or self-host

skip it if

  • you lean on knowledge recall; the lab says Inkling keeps the factuality edge

faq

which is better, Inkling or Inkling Small?

on SWEBench Verified, as each maker reports it, Inkling scores 77.6% and Inkling Small 80.2%. these are each maker's own numbers, not a controlled head-to-head; the spec table lists every benchmark both publish.

which is cheaper, Inkling or Inkling Small?

Inkling Small lists lower input and output prices. on a coding-agent session (2M tokens in, 60,000 out) Inkling costs $2.24 and Inkling Small $1.07.

can i switch between Inkling and Inkling Small mid-task?

yes. in an octomind session, /model octohub:inkling and /model octohub:inkling-small switch models without losing the conversation, so you can draft on one and review on the other.

try either model

switch between models inside the same octomind session.

prices from the catalog of 2026-11-18