thinkingmachines
Thinking Machines: Inkling Small
thinkingmachines/inkling-small
Our take
AI-assisted analysis based on the sources below. Suitability is an editorial assessment; test it on your own workload.
Inkling Small is an open-weight multimodal mixture-of-experts model from Thinking Machines Lab that accepts text, image, and audio inputs and produces text output. The vendor model card describes 276B total parameters with 12B active per token, a 42-layer decoder with hybrid attention, and Apache-2.0 licensing, positioning it as the smaller, more efficient member of its family. The OpenRouter catalog lists a 524,288-token context window, configurable reasoning effort, and tool-use parameters. Vendor-reported benchmarks emphasize agentic and coding results; independent evaluations were not included in this source pack.[1][2]
Strengths
- Open-weight MoE design: sources describe 276B total parameters with only 12B active per token, and the model card releases the weights under the Apache-2.0 license to support research, fine-tuning, and integration.[1][2]
- Native multimodal input: text, images, and audio are encoded into a shared hidden space and processed jointly by the decoder, per the model card.[1][2]
- Configurable operation: the catalog lists a 524,288-token context window, selectable reasoning effort from none to max (high by default), and tool-calling parameters.[1][2]
Limitations
- Evidence is limited to the OpenRouter catalog and the vendor's own model card; all benchmark figures are vendor-reported, so results should be validated on your own workload before adoption.[1][2]
- Recommended input ranges are narrow for best performance: the card suggests image dimensions of 40 to 4096 pixels per side and 16kHz WAV audio ideally under two minutes.[2]
- No knowledge cutoff is reported in the supplied sources, and real-world throughput or quality outside vendor benchmarks is not independently measured here.[1][2]
Where to use it
- Agentic and tool-use systems or coding assistants, which the vendor names as intended uses; treat the reported agentic benchmark results as vendor claims pending your own testing.[2]
- Multimodal applications that combine text, image, and audio in a single request, such as conversational assistants or retrieval-augmented generation pipelines.[1][2]
- Self-hosted or fine-tuned deployments using open weights, with documented recipes for SGLang, vLLM, TokenSpeed, Unsloth, and Hugging Face Transformers.[2]
Sources
- [1] OpenRouter catalog · Thinking Machines: Inkling SmallSources checked: Nov 13, 2026
- [2] Model documentation · Thinking Machines: Inkling SmallSources checked: Nov 13, 2026
About this model
Description below is the original text supplied by OpenRouter, shown in the source language.
Inkling Small is an open-weight multimodal mixture-of-experts model from Thinking Machines Lab, with 12B active parameters out of 276B total. It is positioned as the smaller, more efficient member of...
Specifications
- Model ID
- thinkingmachines/inkling-small
- Canonical ID
- thinkingmachines/inkling-small-20260730
- Creator
- thinkingmachines
- Added to OpenRouter
- 30.07.2026
- Context length
- 524,288
- Modality
- text+image+audio->text
- Input modalities
- text, image, audio
- Output modalities
- text
- Token pricing
- USD per 1M tokens: in $0.450 / out $1.2
- Knowledge cutoff
- —
- Expiration date
- —
Tasks and capabilities
A grade appears only where the source supplies a matching measurement for that task, the model is active and the cohort holds at least 10 comparable models. The capability column lists facts the source declares — supported formats, reasoning and tool parameters — which are not measured quality.
| Task | Grade | Source-declared capability |
|---|---|---|
| General quality | CC — below the B band | reasoning: optional · default high · efforts: max, high, medium, low, minimal, none · parameters: include_reasoning, reasoning, reasoning_effort |
| Coding | BB — next 30% of the cohort | — |
| Agents | BB — next 30% of the cohort | parameters: tool_choice, tools |
| Writing | No measurement | output: text |
| Reasoning | No measurement | reasoning: optional · default high · efforts: max, high, medium, low, minimal, none · parameters: include_reasoning, reasoning, reasoning_effort |
| Document analysis | No measurement | — |
| Vision | No measurement | input: image |
| Image generation | No measurement | — |
| Audio | No measurement | input: audio |
| Video | No measurement | — |
| Embeddings | No measurement | — |
Evaluations
| Source | Benchmark | Value | Grade | Data retrieved |
|---|---|---|---|---|
| Artificial Analysis | Intelligence index | 12.9 | CC — below the B band | 13.11.2026 |
| Artificial Analysis | Coding index | 52.9 | BB — next 30% of the cohort | 13.11.2026 |
| Artificial Analysis | Agentic index | 23.5 | BB — next 30% of the cohort | 13.11.2026 |
How grades work
Each letter grade compares a model only against other models scored by the same source and metric in the current snapshot. S means the top 10%, A the next 20%, B the next 30%, C the rest. Models without a published measurement are shown as not rated — a supported format alone never implies measured quality.
- S — S — top 10% of the cohort
- A — A — next 20% of the cohort
- B — B — next 30% of the cohort
- C — C — below the B band
Providers
Optional live provider statistics from OpenRouter. Missing values are unknown, not zero.
| Provider | Context length | Price in /1M | Price out /1M | Latency p50 (30m, s) | Throughput p50 (30m, tok/s) | Uptime (30m) |
|---|---|---|---|---|---|---|
| DeepInfrafp4 | 524,288 | $0.450 | $1.2 | — | — | 100% |
Provider data retrieved: 13.11.2026
Related articles
No published articles identify this model yet.