Alternatives, in order of how often they win
Every model the official table pits against Inkling 2 Small, ranked by how many shared benchmarks it outscores. A model that beats it often is a genuine substitute; one that rarely does is context.
On the official evaluation table, Kimi K3 outscores Inkling 2 Small on 25 of 64 shared benchmarks, making it the closest substitute by results. Inkling 2 Small is 594B parameters with 17B active and costs $1 per million output tokens on Tinker — cheaper per token than the first-generation models, but slower per task than most competitors.
| Model | Tier | Beats | Trails | Tinker price | Comparison |
|---|---|---|---|---|---|
| Kimi K3 | open, large | 25 | 39 | — | Compare |
| Gemini 3.8 Flash | closed | 23 | 30 | — | Compare |
| DeepSeek-V4.1-Flash-1017 | open, small | 21 | 27 | — | Compare |
| Claude Haiku 5.5 | closed | 21 | 27 | — | Compare |
| GLM 5.3 | open, large | 19 | 35 | — | Compare |
| MiMo V2.6 Pro | open, large | 12 | 30 | — | Compare |
| GLM-5.3-Flash | open, small | 11 | 44 | — | Compare |
| Muse Spark Open | open, large | 8 | 43 | — | Compare |
| MiMo V2.6 Flash | open, small | 6 | 28 | — | Compare |
| Mistral Large 4 | open, large | 6 | 39 | — | Compare |
| GPT-6 Luna | closed | 6 | 41 | — | Compare |
| MiniMax-M3.1-Flash | open, small | 4 | 36 | — | Compare |
How to choose
Raw benchmark count is a blunt instrument. What usually decides it is constraint order: licence first, then context, then price, then score. Inkling 2 Small is the right choice if you need a mid-sized multimodal model, can work within 64K or 256K context, and value fine-tunability — Thinking Machines designed it to adapt quickly to narrow data.
It is the wrong choice if you need the highest possible score, the lowest latency, or audio. Several open-weights models score higher on software engineering, and closed frontier models lead on nearly everything. The model's honest pitch is capability per dollar with a strong vision encoder, not a leaderboard win.
Full matrix · Pricing · Hardware · FAQ