NOUMENOpen app
NOUMEN

thinkingmachines · Open weights

Inkling

Added Jul 15, 2026Open weights

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

Use in the app

The non-thinking version of Thinking Machines' 975B-parameter open-weights Mixture-of-Experts generalist with 41B active parameters. It gives faster direct answers across text, images, and audio, and is built for agentic coding, tool use, detailed instruction following, and long-context work.

  • Tools
  • Vision
  • Structured output
  • Audio input
Input
Text, Images, Audio
Output
Text
Discovery tier
Open weights

Pricing

USD per 1M tokens
Input
$1.00
Output
$4.05
Cache read
$0.17

Reference prices as published by the model provider. API example

Specifications

Context window
1M
Max output
32.8K
Average output (7 days)
596 tokens
Longer than 55% of 465 models

Side by side

Inkling compared

Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.

Compare models

Choose up to 3 models to compare.1 of 3 chosen

Loading models…

InklingthinkingmachinesThis modelTencent Hy3tencentRemove
Providerthinkingmachinestencent
Discovery tierOpen weightsEfficient
AddedJul 15, 2026Jul 6, 2026
Limits
Context window1.05MBest262.14K
Max output32.77K128KBest
Average output596 tokens1.4K tokens
Pricing, USD per 1M tokens
Input$1.00$0.066Best
Output$4.05$0.26Best
Cache read$0.17$0.029Best
Performance
Throughput—73.5 tok/s
Time to first token—4.3 s
Request success (24h)100.0%Best99.5%
Benchmark indexes
Intelligence Index25.025.3Best
Coding Index52.158.8Best
Capabilities
ReasoningNoYes
ToolsYesYes
VisionYesNo
Structured outputYesYes
Audio inputYesNo
Reasoning effort—none, low, high

Public evaluations

Benchmarks

Independent evaluation results, with how each score ranks among the models in the Noumen catalog.

  • Inkling
  • Tencent Hy3

Intelligence Index

  • Inkling25.0
  • Tencent Hy325.3

Inkling tested at xhigh reasoning effort

Inkling: Better than 61% of 326 scored models

3.8Catalog models by score55.1

Indexes

  • Coding Index

    Composite of the coding evaluations

    • Inkling: 52.1
    • Tencent Hy3: 58.8Top

Agentic work

  • AutomationBench-AA

    Workflow automation, penalized for guardrail violations

    • Inkling: 5.0%
    • Tencent Hy3: 18.4%Top

  • AutomationBench-AA tasks completed

    Workflows finished end to end without a guardrail violation

    • Inkling: 0.3%
    • Tencent Hy3: 2.7%Top

  • Harvey LAB-AAv1.1

    Legal tasks passed in full, with no material hallucination

    • Inkling: 0.0%
    • Tencent Hy3: —

  • AA-Briefcasev1.1

    Agentic knowledge work, rated head to head

    • Inkling: 832 Elo
    • Tencent Hy3: 947 EloTop

  • GDPval-AAv2.1

    Economically valuable occupational tasks, rated head to head

    • Inkling: 1,072 EloTop
    • Tencent Hy3: 1,063 Elo

Document reasoning

  • GDP.pdf

    Professional reasoning over PDF documents, all-pass rate

    • Inkling: 12.8%Top
    • Tencent Hy3: 11.2%

  • AA-LCRv1.1

    Reasoning across long documents

    • Inkling: 77.3%
    • Tencent Hy3: 79.0%Top

  • MLCR-AA

    Long-context reasoning over medical records

    • Inkling: 6.1%
    • Tencent Hy3: —

Reasoning and knowledge

  • Humanity’s Last Exam

    Expert-level questions across academic fields

    • Inkling: 31.9%
    • Tencent Hy3: 33.5%Top

  • GPQA Diamond

    Graduate-level science questions

    • Inkling: 87.2%
    • Tencent Hy3: 89.7%Top

Coding

  • Terminal-Benchv4.0

    Practical software engineering in a terminal

    • Inkling: 1.0%Top
    • Tencent Hy3: 0.5%

  • SciCode

    Scientific computing in Python

    • Inkling: 47.0%
    • Tencent Hy3: 48.6%Top

Legacy evaluations

  • AA-LCR (unversioned)

    Long-context reasoning, earlier grading

    • Inkling: 77.3%
    • Tencent Hy3: 79.0%Top

Scores from Artificial Analysis, updated Oct 26, 2026. Inkling was tested at xhigh reasoning effort; compare scores measured at the same effort.Artificial Analysis

Live routing data

Performance and reliability

Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.

Prompt cache hits
0.0%
About $1.00 per 1M input tokens with caching
Request success (24h)
100.0%
Reliable

Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.

Hourly request success · last 72 hoursLowest hour 33.3%
  • Healthy (98% or more)
  • Degraded (90–98%)
  • Down (under 90%)
  • No requests

Developers

Use it through the API

Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.

Read the quickstart
curl https://www.noumen.xyz/api/v1/chat/completions \
  -H "Authorization: Bearer $NOUMEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "thinkingmachines/inkling",
    "messages": [{ "role": "user", "content": "Explain this stack trace." }]
  }'
import os
from openai import OpenAI

client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
    model="thinkingmachines/inkling",
    messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)

thinkingmachines

Other thinkingmachines models, starting with the same family.