The LLM benchmark aggregator.

Inkling

Thinking Machines · 952B · open weights

Strongest in Jailbreak resistance (#2 of 272), weakest in Coding (#128 of 165). Above par in 21 of 35 scopes. Among the models it meets almost everywhere, it finishes behind Claude Fable 5 and Claude Fable 5.1 and ahead of DeepSeek V4 Flash and GPT-6 Luna.

§ 1 · Profile

What it is good at

Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.

Safety59.9−1.6#7/3371/3

Jailbreak resistance64.3−0.6#2/2721/1

Secure code65.4−1.1#9/2741/1

Harm refusal58.7−2.9#19/3001/2

Fairness60.8−9.2#39/3001/2

Toxicity avoidance56.1−2.9#71/2721/1

Knowledge55.3−8.2#46/1381/2

Academic knowledge56.3−7.6#42/1231/1

Factuality54.7−4.3—/0✱✱

Professional53−8.6#51/1681/1

Finance56−6.9#34/1511/1

Medical52.9−13.9#63/1401/1

Legal49.7−17.4#79/1511/1

Human preference61.1−6.4#73/3421/1

Agents52.7−15.4#89/2682/5

Knowledge work53.5−20.1#59/1782/2

Tool use46.2−27.8—/81✱0/1

Web research53.3−10.9—/4✱0/1

Workflow automation45.5−16.5—/0✱✱

Reasoning45.7−21.2#124/1784/4

Science57.7−5.8#40/1221/1

Reasoning46.2−22.9#73/1403/3

Mathematics39.7−26#111/1402/2

Expert reasoning52.3−8.6—/0✱✱

Core abilities47.2−20.2#126/2042/3

Instruction following53.9−19.8#25/571/1

Data analysis49−17.1#37/571/1

Language39.1−32.4#49/571/1

General intelligence46.9−23.2#124/2042/3

Long context56−8.6—/0✱✱

Coding43.4−25.7#128/1652/5

Code generation36.9−33.1#65/771/2

Agentic coding45.2−23.5#106/1572/4

Repository Q&A36−30.7—/0✱✱

§ 2 · Head to head

What it beats, and what beats it

The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for Inkling, left for the other.

Llama 4 Scout Instructwon 23 · lost 0

GPT-4o Miniwon 24 · lost 0

GPT-4.1 Nanowon 24 · lost 1

GPT-5.4 Nanowon 21 · lost 2

GPT OSS 20Bwon 19 · lost 2

Command Awon 20 · lost 3

Gemini 2.0 Flashwon 19 · lost 3

GPT-4owon 18 · lost 3

Llama 4 Maverick Instructwon 21 · lost 4

Nemotron 3 Ultrawon 18 · lost 4

GPT-5.4 Miniwon 18 · lost 4

Gemini 3.5 Flash Litewon 18 · lost 4

DeepSeek V3won 20 · lost 5

Claude Haiku 4.5won 20 · lost 5

Grok 4.3won 22 · lost 6

GPT OSS 120Bwon 18 · lost 6

O3 Miniwon 17 · lost 6

Grok 4.1 Fastwon 16 · lost 6

GPT-4.1 Miniwon 18 · lost 7

GLM-4.6won 17 · lost 7

Kimi K2.5won 17 · lost 7

Gemma 4 31Bwon 17 · lost 8

MiniMax M2.7won 14 · lost 7

Kimi K2won 16 · lost 8

DeepSeek R1won 16 · lost 8

DeepSeek V3.2won 16 · lost 9

GPT-5 Miniwon 15 · lost 10

O4 Miniwon 15 · lost 10

Claude Sonnet 4won 15 · lost 10

Gemini 2.5 Flashwon 15 · lost 10

GPT-5 Nanowon 13 · lost 9

GPT-4.1won 14 · lost 10

Qwen3 235B A22B Instructwon 14 · lost 10

Kimi K2.6won 16 · lost 12

MiniMax M3won 20 · lost 15

Grok 4won 13 · lost 11

Claude 3.7 Sonnetwon 11 · lost 10

Mimo V2.5 Prowon 12 · lost 11

O3won 13 · lost 12

GPT-6 Lunawon 13 · lost 12

DeepSeek V4 Flashwon 16 · lost 15

GLM-5.1won 11 · lost 11

Gemini 2.5 Prowon 12 · lost 13

Claude Opus 4won 11 · lost 12

GLM-5.3 Flashwon 13 · lost 15

Qwen3.6 Pluswon 10 · lost 12

GPT-5.1won 10 · lost 12

Qwen3.7 Maxwon 12 · lost 16

GPT-5won 10 · lost 14

GPT-6 Solwon 10 · lost 15

GPT-5.6 Lunawon 13 · lost 21

DeepSeek V4 Prowon 11 · lost 20

GLM-5.2won 12 · lost 22

GPT-5.2won 9 · lost 19

Gemini 3.5 Flashwon 9 · lost 20

Qwen3.8 27Bwon 8 · lost 18

Claude Sonnet 4.5won 7 · lost 18

Grok 4.5won 8 · lost 21

GPT-5.4won 7 · lost 21

Qwen3.8 Maxwon 7 · lost 21

Gemini 3.6 Flashwon 7 · lost 21

Gemini 3.8 Flashwon 7 · lost 21

Grok 4.7won 6 · lost 19

Claude Sonnet 4.6won 6 · lost 20

GPT-5.6 Solwon 7 · lost 25

GPT-5.5won 6 · lost 22

Muse Spark 1.1won 6 · lost 22

GPT-5.6 Terrawon 7 · lost 26

DeepSeek V4.1 Flashwon 5 · lost 19

Claude Opus 4.6won 5 · lost 19

Grok 4.6won 5 · lost 23

Claude Opus 4.7won 5 · lost 23

Claude Opus 4.5won 4 · lost 19

Gemini 3.1 Prowon 4 · lost 19

Claude Opus 5won 5 · lost 26

Claude Opus 5.5won 4 · lost 21

Claude Sonnet 5won 5 · lost 29

Gemini 3 Prowon 3 · lost 18

Claude Opus 4.8won 4 · lost 24

GLM-5.3won 3 · lost 22

Gemini 3.7 Flashwon 1 · lost 21

Kimi K3won 1 · lost 24

Muse Spark 1.2won 0 · lost 21

Claude Fable 5.1won 0 · lost 22

Claude Fable 5won 0 · lost 23

§ 3 · Sources

Where the numbers come from

10 publications, 53 figures. Every one links to the page it was read from.

LMArena Text ↗1 measureread 2026-09-26

LMArena Text 1442

Artificial Analysis ↗1 measureread 2026-09-26

Artificial Analysis Intelligence Index 25

GDPval-AA ↗1 measureread 2026-09-26

GDPval-AA 1064

AA-Briefcase ↗1 measureread 2026-09-26

AA-Briefcase 834

ARC-AGI-2 ↗1 measureread 2026-09-26

ARC-AGI-2 36.5%

LiveBench ↗8 measuresread 2026-09-26

LiveBench 71.9LiveBench · Reasoning 78.3LiveBench · Coding 71LiveBench · Agentic Coding 49.4LiveBench · Mathematics 88.4LiveBench · Data Analysis 72.8LiveBench · Language 73.5LiveBench · Instruction Following 70.1

SimpleBench ↗1 measureread 2026-09-26

SimpleBench 50%

Vals.ai ↗15 measuresread 2026-09-26

Vals · Legal Research Bench 28.36%Vals · LegalBench 82.95%Vals · Harvey Legal Agent Benchmark 2.08%Vals · Finance Agent 46.6%Vals · CorpFin 68.57%Vals · TaxEval 75.31%Vals · MortgageTax 63.99%Vals · MedCode 41.19%Vals · MedScribe 85.41%Vals · SWE-bench Verified 77.6% (Mini-SWE-agent)Vals · Vibe Code Bench 19.21% (OpenHands)Vals · Code Migration 11.79%Vals · GPQA Diamond 87.12%Vals · MMLU Pro 86.3%Vals · ProofBench 0%

Enkrypt AI Safety Leaderboard ↗6 measuresread 2026-09-26

Enkrypt · Jailbreak risk 0.7%Enkrypt · Harmful content risk 2.2%Enkrypt · CBRN risk 8.7%Enkrypt · Toxicity risk 2%Enkrypt · Bias risk 61%Enkrypt · Insecure code risk 2.2%

✱✱ IFM — K2 Horizon launch ↗the model’s own publisher18 measuresread 2026-09-26

GDPVal-AA 1234tau3-Banking 29.1%Toolathlon Verified 45.5%Automation Bench Public 12.8%Apex-Agents (pass@1) 19%MCPMark 51.2%BrowseComp 77.1%WildClawBench 52.3%Terminal-Bench 2.1 55.1%SciCode 46.1%SWE-Atlas-QnA 25.5%SWE Bench Pro 43.1%Humanity's Last Exam (without tools) 31.9%GPQA Diamond 87.2%CritPt 5.4%AA-LCR 73.3%AA-Omniscience Accuracy 42%AA-Omniscience Non-Hallucination 32%

Badge

[![PublicAI Index](https://publicai.io/model-index/badge?model=inkling)](https://publicai.io/model-index/m/inkling)