Thinking Machines Lab · language model
Inkling: pricing, context and benchmarks
Inkling is a language model from Thinking Machines Lab released on July 17, 2026. It costs $1 per million input tokens and $4.05 per million output tokens, with a 524.3K-token context window. Epoch AI ranks it #65 of 125 rated models.
| Performance (ECI) | 20/40 | |
|---|---|---|
| Price | 9/30 | |
| Context | 10/15 | |
| Features | 12/15 |
What you need to know about Inkling
Inkling was released by Thinking Machines Lab (United States) on July 17, 2026. It accepts text, images, and audio as input and produces text. Its context window reaches 524,288 tokens, larger than 67 % of the models in the registry, and it can write up to 471,859 tokens in one go.
It has been available for about 3 months, long enough for its prices and scores to have settled.
Inkling pricing
On OpenRouter, Inkling costs $1 per million input tokens and $4.05 per million output tokens. Counting three tokens read for every token written, its average price is $1.763 per million, more expensive than 69 % of paid models (2.3 times the median of $0.76).
A concrete example: a task that sends 2,000 tokens (about 1,500 words) and receives a 500-token answer costs $0.004. At 1,000 tasks a day, budget $121 a month. The price calculator redoes this with your own volumes.
Tokens already sent and cached are billed at $0.17 per million.
Performance and benchmarks
The Epoch Capabilities Index (ECI), which aggregates published results across many benchmarks, gives Inkling a score of 148.4 (range 146 to 150.2). That places it #65 of 125 rated models.
Scores published by Epoch AI on the benchmarks the site tracks:
| Benchmark | What it measures | Score | Date |
|---|---|---|---|
| GPQA Diamond | PhD-level science questions | 88.3 % | Jul 15, 2026 |
| ARC-AGI-1 | visual abstract-reasoning puzzles | 79.5 % | Jul 15, 2026 |
| ARC-AGI-2 | harder version of the ARC puzzles | 36.5 % | Jul 15, 2026 |
| FrontierMath | research-level maths problems (tiers 1-3) | 33.3 % | Jul 15, 2026 |
| OTIS Mock AIME | maths competition problems | 88.9 % | Jul 15, 2026 |
| SimpleQA Verified | accuracy on simple facts | 40.3 % | Jul 15, 2026 |
For at least equal performance, Claude Haiku 5.5, GPT-6 Luna, and GPT-5.6 Luna cost at least 30% less.
Features, privacy and local use
- It supports tool calling, so it can drive an AI agent, call APIs or MCP servers.
- It is a reasoning model: it "thinks" before answering, which helps on hard tasks but uses more output tokens.
Its weights are open and downloadable from Hugging Face: you can run it on your own machines without sending your data to a third party.
Other Thinking Machines Lab models
| Model | Released | Avg. price | ECI |
|---|---|---|---|
| Inkling Small | Jul 30, 2026 | $0.64 | 150.3 |
Verdict
Inkling scores 51/100 on our grid, which only uses published figures.
Strengths
- Open weights: self-hostable, your data stays with you.
Avoid if…
- your tasks are mostly simple: a non-reasoning model will cost less.
Alternatives
- GPT-5.4 Mini$1.69
- GPT-5.1$3.44
- Kimi K2.5$1
- Kimi K2.7 Code$1.28
- GPT-5$3.44
- Qwen3.5 397B A17B$1.28
Comparisons and tools
Learn more
Frequently asked questions
How much does Inkling cost?
$1 per million input tokens and $4.05 per million output tokens on OpenRouter (checked October 25, 2026). A task with 2,000 tokens read and 500 written costs $0.004.
How big is Inkling's context window?
524,288 tokens, roughly 393,000 words, with a maximum of 471,859 output tokens.
Can Inkling run locally?
Yes, its weights are published.
Is Inkling suitable for an AI agent?
Technically yes: it supports tool calling. Then check its cost for your volume.
How good is Inkling?
Epoch AI ranks it #65 of 125 rated models (ECI 148.4). On GPQA Diamond, it scores 88.3%.
What are the alternatives to Inkling?
The closest models on price and performance are GPT-5.4 Mini, GPT-5.1, and Kimi K2.5.
Related articles
- Mistral AI: models, pricing and agentsOctober 25, 2026
- LLM benchmarks: how to read model rankingsOctober 24, 2026
- Running a local LLM: hardware, software, and limitationsOctober 24, 2026
- Open-source or closed models for your AI agents?October 5, 2026
Neutrality. No affiliate links. No vendor pays to be reviewed, rated or ranked. Reviews are directed by François Chollet, neutral and factual, and every data point is dated and sourced. How we score