Inkling-Small

Name: Inkling-Small
Author: Thinking Machines Lab

Multimodal

Thinking Machines Lab

Inkling-Small is Thinking Machines Lab's efficient open-weights MoE multimodal model (276B total / 12B active) released under Apache 2.0. It accepts text, image, and audio inputs and generates text, with native reasoning, variable thinking effort, and a context window up to 1M tokens. It scores 80.2% on SWE-Bench Verified and is served through Tinker at $0.30 / $1.20 per 1M input/output tokens.

Key Specifications

Parameters

276.0B

Context

256.0K

Release Date

July 30, 2026

Average Score

60.3%

API Documentation Repository Model Weights Results Blog

Timeline

Key dates in the model's history

Announcement

July 30, 2026

Last Update

August 31, 2026

Technical Specifications

Parameters

276.0B

Training Tokens

Knowledge Cutoff

Family

Capabilities

MultimodalZeroEval

Pricing & Availability

Input (per 1M tokens)

$0.30

Output (per 1M tokens)

$1.20

Max Input Tokens

256.0K

Max Output Tokens

256.0K

Supported Features

Function CallingStructured OutputCode ExecutionWeb SearchBatch InferenceFine-tuning

Benchmark Results

Model performance metrics across various tests and benchmarks

Programming

Programming skills tests

SWE-Bench Verified

Bash-only harness (same as Inkling 77.6%) • Self-reported

80.2%

Reasoning

Logical reasoning and analysis

GPQA

GPQA Diamond • Self-reported

89.5%

Other Tests

Specialized benchmarks

AA-Briefcase

Elo score; Artificial Analysis (max_score 3000) • Self-reported

30.6%

AIME 2026

• Self-reported

95.5%

ARC-AGI

ARC-AGI-1 • Self-reported

84.0%

ARC-AGI v2

ARC-AGI-2 • Self-reported

40.1%

Artificial Analysis

Artificial Analysis Intelligence Index v4.1. Score 40 (vendor model card). • Self-reported

40.0%

BrowseComp

With context management • Self-reported

77.4%

CharXiv-R

CharXiv RQ original • Self-reported

77.4%

CritPT

• Self-reported

8.3%

GDPval-AA

GDPval-AA v2 Elo (max_score 3000) • Self-reported

42.3%

Global-MMLU-Lite

• Self-reported

86.7%

Humanity's Last Exam

Text only • Self-reported

31.6%

IFBench

• Self-reported

82.2%

MCP Atlas

Public set • Self-reported

79.6%

MMAU

• Self-reported

77.0%

MMMU-Pro

MMMU Pro Standard 10 • Self-reported

74.0%

SciCode

• Self-reported

48.7%

SimpleQA Verified

• Self-reported

20.6%

SWE-Bench Pro

SWE-Bench Pro public • Self-reported

55.9%

Tau3 Banking

• Self-reported

15.5%

Terminal-Bench 2.1

Vendor best harness (model card) • Self-reported

64.7%

Toolathlon

Toolathlon Verified • Self-reported

54.4%

VoiceBench Avg

• Self-reported

90.1%

License & Metadata

License

apache_2_0

Announcement Date

July 30, 2026

Last Updated

August 31, 2026

Compare Inkling-Small

All comparisons

vs Claude Sonnet 5 vs Claude Fable 5 vs GPT-5.6 Terra vs Seed 2.1 Turbo vs GLM-5.3 Flash vs Grok 4.6 vs Hy4 preview vs Laguna S 2.1

Similar Models

All Models

Command A+

Cohere

MM218.0B

Best score:0.8 (TAU)

Released:May 2026

Step-3.5-Flash

StepFun

MM196.0B

Best score:0.9 (TAU)

Released:Feb 2026

Price:$0.10/1M tokens

Kimi K3

Moonshot AI

MM2.8T

Best score:0.9 (GPQA)

Released:Jul 2026

Price:$3.00/1M tokens

Kimi K2.5

Moonshot AI

MM1.0T

Best score:0.9 (GPQA)

Released:Jan 2026

Qwen3.8 Flash Next

Alibaba

MM180.0B

Best score:0.9 (GPQA)

Released:Aug 2026

Price:$0.16/1M tokens

Qwen3.8 Flash

Alibaba

MM125.0B

Best score:0.9 (GPQA)

Released:Aug 2026

Price:$0.15/1M tokens

GLM-4.6

Zhipu AI

MM357.0B

Best score:0.8 (GPQA)

Released:Sep 2025

Price:$0.60/1M tokens

MiniMax M2.5

MiniMax

MM230.0B

Released:Feb 2026

Recommendations are based on similarity of characteristics: developer organization, multimodality, parameter size, and benchmark performance. Choose a model to compare or go to the full catalog to browse all available AI models.