Skip to main content
Thinking Machines

Inkling 2 Mini

Inkling 2 Mini is a chat model from Thinking Machines, positioned as the smaller "Mini" variant in the Inkling 2 line.

Input from
$0.100 / 1M tokens
on OpenRouter · 2 providers
Output from
$0.400 / 1M tokens
on OpenRouter

API Pricing

Cheapest input on OpenRouter · cheapest output on OpenRouter
ProviderInput / 1MOutput / 1MCached / 1M
$0.100$0.400$0.020
$0.100$0.400$0.020

Prices updated daily. Last check: Nov 21, 2026

Inkling 2 Mini pricing by provider

Input, output, and batch rates, plus alternatives, for one provider at a time.

Performance & Benchmarks

Source: Artificial Analysis →
Intelligence
20.1 / 100
Output Speed
261 t/s
Latency (TTFT)
1.6s

Benchmarks measured Nov 2026. Scores are independent evaluations, not vendor-reported.

Model Details

General

Creator
Thinking Machines
Modalities
Text

Capabilities

Strengths & Limitations

Strengths

  • Measured output throughput of approximately 261 tokens per second (Artificial Analysis), fast enough for smooth streamed chat output
  • "Mini" tier of the Inkling 2 family, aimed at higher-volume or latency-sensitive deployments rather than maximum-capability workloads
  • Available through hosted inference APIs, so no local GPU provisioning is required to evaluate it
  • Chat-formatted interface, making it a drop-in candidate for existing assistant and conversational pipelines
  • Speed profile is independently measured by a third party rather than vendor-reported
  • Multiple-provider tracking on this page allows direct comparison of serving cost and configuration

Limitations

  • Time to first token of roughly 1.55 seconds is slower than the fastest small chat models, which affects very short interactive turns
  • We do not currently have a confirmed context window figure for this model
  • No public benchmark scores for reasoning, coding, or math are tracked in our database for Inkling 2 Mini
  • Modality support beyond text chat is not confirmed in our data, so image or audio input should be verified with the provider
  • As a Mini-tier model, it is generally not the Inkling 2 variant intended for the hardest multi-step tasks
  • Provider availability for Inkling models is narrower than for widely distributed open-weight families

Key Features

•Chat-completion interface for multi-turn conversation
•Mini tier within the Inkling 2 model family from Thinking Machines
•Measured output speed of ~260.8 tokens per second (Artificial Analysis)
•Measured time to first token of ~1,550 ms (Artificial Analysis)
•Streaming token output suitable for live assistant UIs
•Hosted API access through tracked inference providers
•Per-provider pricing comparison available in the table on this page

About Inkling 2 Mini

Inkling 2 Mini is a conversational (chat) language model released by Thinking Machines. The "Mini" designation places it as the compact member of the Inkling 2 family, the tier typically selected when request volume, latency, or cost matter more than maximum capability on hard reasoning tasks. It is served through hosted inference APIs, and availability varies by provider — the pricing table on this page lists the providers we currently track for it. Independent measurements collected from Artificial Analysis put Inkling 2 Mini at roughly 261 output tokens per second with a time to first token of about 1.55 seconds. The throughput figure is in the range where streamed responses render faster than most users read, which suits interactive chat and assistant interfaces; the first-token latency is less aggressive, so very short request/response turns will feel dominated by that initial delay rather than by generation speed. Both numbers are provider- and load-dependent and should be treated as a snapshot rather than a guarantee. We do not currently track a confirmed context window, modality list, or public benchmark scores for Inkling 2 Mini, so this page intentionally does not characterize its reasoning, coding, or vision behavior. Readers evaluating it against other small chat models should test it directly on their own prompts and compare the measured speed figures above alongside the live per-provider pricing below.

Common Use Cases

Inkling 2 Mini fits workloads where a compact chat model is run at volume: customer-facing assistants, in-product chat widgets, summarization and rewriting passes over moderate amounts of text, drafting and reformatting, conversational front-ends that route harder queries to a larger model, and batch classification or extraction jobs expressed as chat prompts. Its roughly 261 tokens/second output rate means long generated answers stream quickly, which matters for assistant UIs where users watch text appear. The ~1.55 second time to first token makes it a weaker fit for ultra-low-latency uses such as autocomplete or voice turn-taking, where the initial delay dominates perceived responsiveness. Because we do not track confirmed benchmark scores or a context window for this model, teams considering it for long-document or complex reasoning work should validate those properties against their own test set before committing.

Frequently Asked Questions

How much does Inkling 2 Mini cost?

Pricing is set by each inference provider and differs by pricing type — input tokens, output tokens, cached input, and any batch or committed-throughput discounts are all priced separately, and rates change frequently. Check the pricing table on this page for the current per-provider figures we track.

What is Inkling 2 Mini best used for?

It is aimed at high-volume chat workloads: product assistants, summarization and rewriting, drafting, extraction and classification framed as chat prompts, and as the cheap first tier in a routing setup that escalates difficult queries to a larger model. Its measured ~261 tokens/second output rate suits interfaces that stream responses to users.

How fast is Inkling 2 Mini?

Third-party measurements from Artificial Analysis report about 260.8 output tokens per second with a time to first token of roughly 1,550 milliseconds. Generation speed is strong for a small chat model; the first-token latency is mid-range, so short exchanges will feel gated by that startup delay. Both figures vary with provider, region, and load.

How does Inkling 2 Mini differ from other Inkling 2 variants?

The "Mini" name marks it as the compact tier of the Inkling 2 family from Thinking Machines, which generally means lower serving cost and faster generation in exchange for less headroom on difficult multi-step tasks than larger siblings. We do not track comparative benchmark scores across the family, so direct testing on your own workload is the reliable way to decide.

What context window does Inkling 2 Mini support?

We do not have a confirmed context window figure for Inkling 2 Mini in our database. If your workload depends on long inputs, verify the supported context length with the specific provider you plan to use, since serving limits can also be set below a model's maximum.