Speka
Start free
All models
reasoningThinking Machines

Inkling

Auto-discovered from the NVIDIA catalog on 2026-07-28.

thinkingmachines/inkling
Running· verified just now · 642ms

What is Inkling?

Inkling is Thinking Machines's reasoning model. Auto-discovered from the NVIDIA catalog on 2026-07-28. It runs on Speka's OpenAI-compatible /v1/chat/completions endpoint at $0.20 / 1M input and $0.20 / 1M output tokens, using the model ID thinkingmachines/inkling.

Test it live

Live playgroundInklingin-browser demo

Test Inkling live

Send a message and watch the model respond. Multi-turn — it remembers the conversation.

How to call Inkling via the API

Speka exposes Inkling on an OpenAI-compatible /v1/chat/completions endpoint. Point any OpenAI SDK at https://speka.me/v1, set model to thinkingmachines/inkling, and your existing code runs unchanged.

quickstart.py
from openai import OpenAI

client = OpenAI(
    base_url="https://speka.me/v1",
    api_key="sk-speka-live-...",  # your Speka key
)

resp = client.chat.completions.create(
    model="thinkingmachines/inkling",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(resp.choices[0].message.content)
FAQ

Frequently asked questions

Inkling is Thinking Machines's reasoning model. Auto-discovered from the NVIDIA catalog on 2026-07-28. It is available on Speka's OpenAI-compatible API using the model ID thinkingmachines/inkling.
On Speka, Inkling is priced at $0.20 / 1M input and $0.20 / 1M output tokens. Pricing is per-token with no subscription or seat fee — usage is billed against your plan's monthly allowance, then at the same rate with no overage penalty. Rates are current as of June 2026; check the pricing page for updates.
Speka exposes Inkling on an OpenAI-compatible /v1/chat/completions endpoint. Set base_url to https://speka.me/v1, supply your Speka key, and set model to "thinkingmachines/inkling" in any OpenAI SDK — no other code changes are required.
Yes. Inkling is served through Speka's OpenAI-compatible endpoint, so any application built on the OpenAI Python or Node.js SDK, LangChain, or LlamaIndex can call it by changing only the base URL and model ID.

More reasoning models

DeepSeek logo

DeepSeek V4 Flash

DeepSeek
reasoning

Fast open reasoning model with transparent chain-of-thought. Strong at math, logic and multi-step problem solving, with low latency suitable for production and automation.

deepseek-ai/deepseek-v4-flash
Input / 1M
$0.27
Output / 1M
$1.10
Open model
DeepSeek logo

DeepSeek V4 Pro

DeepSeek
reasoning

The larger DeepSeek V4 tier: deeper multi-step reasoning, stronger maths and code, for tasks where answer quality matters more than raw speed.

deepseek-ai/deepseek-v4-pro
Input / 1M
$0.55
Output / 1M
$2.20
Open model
NVIDIA logo

Nemotron 3 Super 120B

NVIDIA
reasoning

NVIDIA's mixture-of-experts reasoning model — 120B total parameters with 12B active, so frontier-class reasoning arrives at small-model latency.

nvidia/nemotron-3-super-120b-a12b
Input / 1M
$0.40
Output / 1M
$0.60
Open model