Inkling Small
Inkling-Small is a lighter-weight model with 12B active parameters, trained with a similar recipe, to Inkling that achieves strong performance with even lower cost and latency.
text + image + file → text · made by Thinking Machines Lab
Free
Offered at no charge, which is not the same as cheap: a free lane carries a rate limit and can be withdrawn. The prices above are what you pay when it is not available to you.
- OpenRouter
Sold by 20 ways
| Seller | Lane | Rate |
|---|---|---|
| Thinking Machines Lab api | standard | $0.3per Mtok in$1.2per Mtok out$0.06per Mtok cachedtinker-docs.thinkingmachines.ai · read 2026-08-25 |
| Thinking Machines Lab api | training | $0.58per Mtok in$1.44per Mtok out$0.12per Mtok cachedtinker-docs.thinkingmachines.ai · read 2026-08-25 |
| Thinking Machines Lab api | training-256k | $1.16per Mtok in$2.89per Mtok out$0.23per Mtok cachedtinker-docs.thinkingmachines.ai · read 2026-08-25 |
| OpenRouter 2 ways | $0.45per Mtok in$1.2per Mtok out$0.1per Mtok cached | |
| OpenRouter aggregator | free | $0per Mtok in$0per Mtok outopenrouter.ai · read 2026-10-31 |
| OpenRouter aggregator | standard | $0.45per Mtok in$1.2per Mtok out$0.1per Mtok cachedopenrouter.ai · read 2026-08-24 |
| OpenRouter aggregatornot on the seller's list today | baseten/fp8 | $0.5per Mtok in$1.2per Mtok out$0.1per Mtok cachedopenrouter.ai · read 2026-08-25 |
| OpenRouter aggregatornot on the seller's list today | batch | $0.5per Mtok in$1.2per Mtok out$0.1per Mtok cachedopenrouter.ai · read 2026-08-29 |
| OpenRouter aggregatornot on the seller's list today | together | $0.5per Mtok in$1.2per Mtok out$0.1per Mtok cachedopenrouter.ai · read 2026-08-25 |
| Nous Research aggregatornot on the seller's list today | resold | $0.36per Mtok in$0.96per Mtok out$0.08per Mtok cachedinference-api.nousresearch.com · read 2026-08-25 |
| Nous Research aggregator | standard | $0.45per Mtok in$1.2per Mtok out$0.1per Mtok cachedinference-api.nousresearch.com · read 2026-09-12 |
| Nous Research aggregatornot on the seller's list today | batch | $0.5per Mtok in$1.2per Mtok out$0.1per Mtok cachedinference-api.nousresearch.com · read 2026-09-12 |
| DeepInfra aggregator | standard | $0.45per Mtok in$1.2per Mtok outapi.deepinfra.com · read 2026-08-25 |
| ElectronHub 0 ways | $0.45per Mtok in$1.2per Mtok out | |
| ElectronHub aggregatornot on the seller's list today | standard | $0.45per Mtok in$1.2per Mtok outapi.electronhub.ai · read 2026-09-05 |
| Hugging Face Inference Providers 2 ways | $0.5per Mtok in$1.2per Mtok out | |
| Hugging Face Inference Providers aggregator | deepinfra | $0.45per Mtok in$1.2per Mtok outrouter.huggingface.co · read 2026-08-25 |
| Hugging Face Inference Providers aggregator | standard | $0.5per Mtok in$1.2per Mtok outmodels.dev · read 2026-10-31 |
| Hugging Face Inference Providers aggregatornot on the seller's list today | baseten | $0.5per Mtok in$1.2per Mtok outrouter.huggingface.co · read 2026-08-25 |
| Hugging Face Inference Providers aggregatornot on the seller's list today | together | $0.5per Mtok in$1.2per Mtok outrouter.huggingface.co · read 2026-08-25 |
| Vercel AI Gateway aggregator | standard | $0.45per Mtok in$1.2per Mtok out$0.1per Mtok cachedai-gateway.vercel.sh · read 2026-09-26 |
| Baseten aggregator | standard | $0.5per Mtok in$1.2per Mtok out$0.1per Mtok cachedraw.githubusercontent.com · read 2026-10-31 |
| Together AI 0 ways | $0.5per Mtok in$1.2per Mtok out$0.1per Mtok cached | |
| Together AI aggregatornot on the seller's list today | standard | $0.5per Mtok in$1.2per Mtok out$0.1per Mtok cacheddocs.together.ai · read 2026-08-24 |
Measured 19 standings
| Place | Board | Metric | Score |
|---|---|---|---|
| 56thof 131 | Artificial Analysis · Agentic Index | Index | 23.5 |
| 58thof 122 | Frontiermath tiers 1 3 v2 — Epoch AI | mean_score (xhigh) | 46.316 |
| 60thof 78 | Frontiermath tier 4 v2 — Epoch AI | mean_score (xhigh) | 17.073 |
| 63rdof 326 | GPQA Diamond | mean_score (xhigh) | 88.51 |
| 74thof 163 | Artificial Analysis · Coding Index | Index | 52.9 |
| 74thof 86 | Artificial Analysis · Intelligence Index | Artificial Analysis Intelligence Index | 13 |
| 75thof 84 | Proofbench — Epoch AI | Accuracy | 6 |
| 78thof 306 | OTIS Mock AIME 2024-2025 | mean_score (xhigh) | 90 |
| 81stof 93 | Simpleqa verified — Epoch AI | mean_score (xhigh) | 19.1 |
| 85thof 135 | Webdev arena — Epoch AI | Arena Score | 1406.58 |
| 110thof 273 | ARC-AGI-1 (Public Eval) | Score (%) (xhigh) | 89.38 |
| 112thof 282 | ARC-AGI-2 (Public Eval) | Score (%) (xhigh) | 39.31 |
| 118thof 298 | ARC-AGI-1 (Semi-Private) | Score (%) (xhigh) | 84 |
| 122ndof 298 | ARC-AGI-2 (Semi-Private) | Score (%) (xhigh) | 40.14 |
| 124thof 235 | Scicode — Epoch AI | Score | 49.653 |
| 127thof 260 | Critpt — Epoch AI | Accuracy | 8.286 |
| 129thof 290 | Arc agi 2 — Epoch AI | Score (xhigh) | 40.139 |
| 129thof 309 | Arc agi — Epoch AI | Score (xhigh) | 84 |
| 132ndof 142 | Mystery game puzzles — Epoch AI | mean_score (xhigh) | 6 |
About
Inkling Small — a text model from Thinking Machines Lab, sold by 7 companies from $0.3 in and $1.2 out per million tokens, placed 63rd of 326 on GPQA Diamond.
It takes text, images and files and returns text, with a context window of 1,000,000 tokens. It was published in July 2026. Its sellers say it can reason step by step and call a tool. The catalogue files it under reasoning. Seven companies sell it. The cheapest is $0.3 in and $1.2 out per million tokens at Thinking Machines Lab. Beside the standard rate there is a separately routed lane. It has been measured on 19 boards, and stands best at 63rd of 326 on GPQA Diamond.
Every current figure
- Maker
- Thinking Machines Lab
- Register
- model
- Takes
- text + image + file
- Returns
- text
- Context
- 1,000,000 tokens
- Published
- July 2026
- Parameters
- 266 billion
- Licence
- not read
- Sellers
- 7
- Price
- $0.3 in and $1.2 out per million tokens — Thinking Machines Lab
- Boards
- 19
- Best place
- 63rd of 326 — GPQA Diamond
Known as 3 names
thinkingmachines/inkling-smallthinkingmachines/Inkling-Small:peft:262144thinkingmachines/Inkling-Small:peft:262144:sampling-nvfp4