Together AI serves Inkling 2 Small with a 1M-token context at $0.30 input and $1.00 output (November 12, 2026)
Together AI added thinkingmachines/Inkling-2-Small on serverless: 1,048,576 context, image input, $0.30 in and $1.00 out per 1M tokens.
Your keys
0 of 1 ready
Create each key on its own page, paste it below, then copy the finished lines into your project's .env file. What you paste stays in this tab. It is never sent to Wirestack or saved.
- Together AI API keyTOGETHER_API_KEYGet key
Select Create key on your project's API keys page. The key is shown only once.
What changed
On November 12, 2026, Together AI added thinkingmachines/Inkling-2-Small to its serverless models. The changelog lists:
- Context length: 1,048,576 tokens
- Quantization: NVFP4
- Input: text and image
- Function calling: supported
- Structured outputs: supported
- Price per 1M tokens: $0.30 input, $1.00 output, $0.03 cached input
The serverless models page shows the same model ID, context length and prices, and lists the model in both the chat and the vision tables. The together.ai pricing page shows the same three prices.
What to do
Nothing changes for existing integrations. To try the model, set model to thinkingmachines/Inkling-2-Small in a request you already send to Together. You need a Together API key in TOGETHER_API_KEY.
On the bill: the Kimi K3.1 Pro model that Together added on November 5 has the same context length at $2.00 input and $10.00 output per 1M tokens, so Inkling 2 Small is a much cheaper way to send very long prompts. Cached input is billed at $0.03 per 1M tokens, a tenth of the normal input price.
Function calling and structured outputs are both marked as supported, so an integration that uses tools or a JSON schema can switch the model ID without removing those parts.
Not confirmed
The changelog does not say whether the model works through the Messages API and Responses API that Together added on November 3, or only through chat completions. It gives no rate limits and does not mention dedicated inference or post-training for this model.
