Inkling is a large language model from Thinking Machines Lab, released on July 15, 2026. It accepts text, image and audio input.

When was Inkling released?

Thinking Machines Lab released Inkling on July 15, 2026.

How much does Inkling cost?

Inkling costs $1 per 1M input tokens and $4.05 per 1M output tokens through the Thinking Machines Lab API. Cached input costs $0.17 per 1M tokens. Prices are from Thinking Machines Lab's official pricing as of October 5, 2026.

What is Inkling's context window?

Inkling has a 1M-token context window.

How does Inkling score on benchmarks?

Inkling has 39 cited benchmark scores, including APEX-Agents (Mercor) 33.8%, Chess Puzzles (Epoch AI run) 21% and FrontierMath Tier 4 v2 (Epoch AI run) 4.9%. Each score links to its source.

Inkling

Thinking Machines Lab·Pending·Released 2026-07-15textvisionaudio

Context1,000k

Max output-

Input / 1M$1 ↗

Output / 1M$4.05 ↗

Cached in$0.17

Benchmarks39

Benchmarks

Cited public sources · rank is exact-benchmark, apples-to-apples

APEX-Agents (Mercor)

#42/4933.8% ↗

Chess Puzzles (Epoch AI run)

#50/9221% ↗

FrontierMath Tier 4 v2 (Epoch AI run)

#51/574.9% ↗

FrontierMath Tiers 1-3 v2 (Epoch AI run)

#56/7333.3% ↗

FrontierSWE V2 (Proximal Labs)

#18/184.1% ↗

GPQA Diamond (Epoch AI run)

#45/10288.3% ↗

LiveBench Agentic Coding (LiveBench)

#35/5649.4% ↗

LiveBench Coding (LiveBench)

#50/5671% ↗

LiveBench Data Analysis (LiveBench)

#36/5672.8% ↗

LiveBench Instruction Following (LiveBench)

#26/5670.1% ↗

LiveBench Language (LiveBench)

#49/5673.5% ↗

LiveBench Mathematics (LiveBench)

#35/5688.4% ↗

LiveBench Reasoning (LiveBench)

#46/5678.3% ↗

OTIS Mock AIME 2024-2025 (Epoch AI run)

#48/10388.9% ↗

SAGE (Vals AI)

#59/7736.6% ↗

SimpleQA Verified (Epoch AI run)

#43/7140.3% ↗

Terminal-Bench 4.0 (Vals AI)

#37/400.5% ↗

WeirdML (Håvard Tveit Ihle)

#87/9232.3% ↗

LMArena Agent (LMArena)

#46/48-0.1086 ↗

LMArena Elo

#66/1411441.4 ↗

LMArena WebDev (LMArena)

#58/941412.6 ↗

tau2-bench Banking Knowledge (Sierra)

#18/2225% ↗

SimpleBench (SimpleBench)

#49/7350% ↗

AIME 2026

#1/597.1% ↗

Audio MC

56.6% ↗

BrowseComp (w/ ctx management)

77.1% ↗

CharXiv RQ

#1/278.1% ↗

CharXiv RQ (with python)

#1/282% ↗

Global-MMLU-Lite

88.7% ↗

IFBench

#4/1279.8% ↗

MCP Atlas

#5/1376% ↗

MMAU

#1/277.2% ↗

SimpleQA Verified

#3/843.9% ↗

SWE-bench Pro (public)

#2/254.3% ↗

SWE-bench Verified

#14/5977.6% ↗

Terminal-Bench 2.1 (best harness)

#2/263.8% ↗

Toolathlon Verified

#6/645.5% ↗

Toolathlon-Verified (HKUST)

#20/2345.5% ↗

VoiceBench

91.4% ↗

Compare

vs Gemini 4 Argon vs GPT-6.1 Sol vs Claude Sonnet 5.5 vs Muse Spark 1.3 vs GPT-6 Sol vs GPT-6 Luna vs Claude Opus 5.5 vs GPT-6 Astra

More from Thinking Machines Lab

Inkling-Small

Inkling FAQ

What is Inkling?: Inkling is a large language model from Thinking Machines Lab, released on July 15, 2026. It accepts text, image and audio input.
When was Inkling released?: Thinking Machines Lab released Inkling on July 15, 2026.
How much does Inkling cost?: Inkling costs $1 per 1M input tokens and $4.05 per 1M output tokens through the Thinking Machines Lab API. Cached input costs $0.17 per 1M tokens. Prices are from Thinking Machines Lab's official pricing as of October 5, 2026.
What is Inkling's context window?: Inkling has a 1M-token context window.
How does Inkling score on benchmarks?: Inkling has 39 cited benchmark scores, including APEX-Agents (Mercor) 33.8%, Chess Puzzles (Epoch AI run) 21% and FrontierMath Tier 4 v2 (Epoch AI run) 4.9%. Each score links to its source.

Access Inkling and every other model through one endpoint with automatic failover: Respan gateway.