Inkling — benchmark results
Thinking Machines Lab's first open-weights model, a 975B-A41B multimodal MoE with controllable thinking effort. Provider: Thinking Machines. Released 2026-07-15. Access: Open.
Unified ELO 1756 ± 20, rank #186 of 1776 rated models, from 43 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Vals AI CorpFin v2 | 68.57 | Accuracy (%) | 97.5 |
| AI for Education Pedagogy - Social studies | 87.27 | Accuracy (%) | 93.2 |
| AI for Education Pedagogy - Technology | 86.79 | Accuracy (%) | 93 |
| AI for Education SEND | 83.03 | Accuracy (%) | 91.5 |
| BenchLM | 67.5 | Overall Score | 90.4 |
| AI for Education Pedagogy - Secondary | 87.42 | Accuracy (%) | 89.8 |
| Vals AI TaxEval v2 | 75.31 | Accuracy (%) | 89.5 |
| AI for Education Pedagogy | 87.65 | Accuracy (%) | 85.5 |
| AI for Education Pedagogy - Science | 90.71 | Accuracy (%) | 85 |
| Vals AI MedScribe | 85.41 | Accuracy (%) | 84.7 |
| Chatbot Arena (Text) | 1445 | Elo | 83.9 |
| Vals AI LiveCodeBench | 85.52 | Accuracy (%) | 83.5 |
Interactive version: theaggregate.ai/model?slug=inkling · How the rankings work · Data refreshed daily, snapshot 2026-07-22.