Thinking Machines Lab models, access and benchmark guide
Thinking Machines Lab model and access guide with a practical US-English evaluation brief and links to publisher materials.

👁 111 reads
A small model can earn its place by routing easy, repetitive messages. An Inkling result on a broad benchmark does not prove it can distinguish a library hold from “hold off.” Give the evaluator real-looking English counterexamples with known labels.
The Inkling 2 Small result concerns a general intelligence evaluation. A library classifier has a narrower job. Give it a copied email thread in which the older message asks for a hold and the new reply cancels that request. Compare the action it proposes with what the last sender actually wrote. For a purchase, require the deployment name along with the quoted rate. A product branded Thinking Machines can expose a different model and still produce superficially similar results.
Choosing the first job
Route invented renewals and cancellations, reserving an explicit undecided label for genuinely ambiguous messages.
Models in the November 17 snapshot
| Model and track | Placement | Reported score |
|---|---|---|
| Inkling 2 Small (AA intelligence) | 20 | 31 index points |
| Inkling 2 Small (AA additions) | ||
| Inkling 2 Small (Arena WebDev) | 24 | 1645 points |
| Inkling 2 Small (Arena agents) | 23 | 3.32%±1.79% |
Release record. Publisher documentation identifies the product family. The ranked configurations include Inkling 2 Small.
Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →Sources
- https://thinkingmachines.ai/ (accessed November 17, 2026)
- https://thinkingmachines.ai/blog/ (accessed November 17, 2026)
- https://artificialanalysis.ai/leaderboards/models (accessed November 17, 2026)
- https://artificialanalysis.ai/changelog (accessed November 17, 2026)
- https://arena.ai/leaderboard/code (accessed November 17, 2026)
- https://arena.ai/leaderboard/agent (accessed November 17, 2026)
- https://artificialanalysis.ai/models/inkling-2-small (accessed November 17, 2026)
More on this from QUANTSLANT
ai tools·Text-to-speech voices ranked for English narrationCompare provider-voice speech rankings and use a local English script to check names, pacing and pronunciation.