Inkling

Inkling ranks first on 2 of the 138 benchmarks with five or more models scored.

Thinking Machines Released July 15th, 2026 Compare with others

140

Benchmarks with a score

Ranked first, out of five or more models

In the top three

Thinking Machines Jul 2026

Inkling

Open weights

148.4 #76 of 256

Epoch Capabilities Index

Self-Reported Index 150.1 #91 of 404

MMMU-Pro 73.5% #47 /90
IFBench 79.8% #9 /80
Terminal-Bench 4.0 0% #40 /40

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

Benchmark	Score	Rank	Source
IFBench	79.8%	#9 of 80	Lab model card
LMArena (Elo)	1,443	#71 of 335	Official leaderboard
Longform Creative Writing (EQ-Bench)	72.5	#35 of 119	Official leaderboard
Humanity's Last Exam	46%	#54 of 154	Lab model card
Creative Writing v3	70.9	#49 of 120	Official leaderboard
SEA-HELM	71.3	#31 of 71	Official leaderboard
EQ-Bench 4	1,212	#16 of 32	Official leaderboard
AA-Omniscience	2.1	#5 of 10	Lab model card
Global MMLU Lite 3.0	88.7%	#2 of 4	Lab model card
SimpleQA Verified	40.3%	#56 of 94	Epoch AI run
AA-LCR Extended	50.6%	#34 of 49	Official leaderboard
StudyArena Text	1,190	#68 of 85	Official leaderboard
DesignArena ASCII Art	1,087	#70 of 77	Official leaderboard

Math and science

Benchmark	Score	Rank	Source
AIME 2026	97.1%	#5 of 38	Lab model card
GPQA Diamond	88.3%	#61 of 229	Epoch AI run
OTIS Mock AIME 2024-2025	88.9%	#74 of 212	Epoch AI run
DTBench	87.5%	#65 of 180	Official leaderboard
ARC-AGI-2	36.5%	#48 of 98	Official leaderboard
ARC-AGI	79.5%	#49 of 95	Official leaderboard
SimpleBench	50%	#60 of 106	Official leaderboard
FrontierMath Tiers 1-3 v2	33.3%	#66 of 86	Official leaderboard
FrontierMath Tier 4 v2	4.88%	#60 of 68	Official leaderboard
ProofBench	2.00%	#57 of 59	Official leaderboard
ProofBench v1.1	0%	#51 of 52	Official leaderboard
Prophet Arena	0.16	#1 of 1	Lab model card

Coding

Benchmark	Score	Rank	Source
SWE-bench Verified	77.6%	#33 of 155	Lab model card
DesignArena Website	1,216	#73 of 151	Official leaderboard
DesignArena Code	1,195	#76 of 143	Official leaderboard
DesignArena Agentic Overall	1,165	#40 of 74	Official leaderboard
DesignArena UI Component	1,190	#76 of 139	Official leaderboard
DesignArena 3D Design	1,145	#76 of 136	Official leaderboard
DesignArena Game Dev	1,153	#85 of 143	Official leaderboard
DesignArena Frontend Web App	1,177	#32 of 53	Official leaderboard
WebDev Arena	1,410	#64 of 105	arena.ai, via Epoch AI
Code Arena	1,414	#81 of 128	Official leaderboard
SWE-Bench Pro	54.3%	#53 of 81	Lab model card
ALE-Bench	1,195	#83 of 126	Official leaderboard
DesignArena SVG	1,102	#73 of 103	Official leaderboard
DesignArena Data Visualization	1,158	#87 of 118	Official leaderboard
Vibe Code Bench	19.2%	#91 of 120	Official leaderboard
Code Migration	11.8%	#63 of 78	Official leaderboard
SWE Atlas Codebase QnA Extended	1.30%	#32 of 38	Official leaderboard
SWE-bench Verified Extended	26.7%	#38 of 44	Official leaderboard
WeirdML	32.3%	#97 of 112	Official leaderboard
Real-SWE 1.1	8.00%	#13 of 15	Official leaderboard
SWE-Bench Pro V2 Hard	56.9%	#16 of 18	Official leaderboard
APEX-SWE	26.1%	#58 of 65	Official leaderboard
SWE-Bench Pro V2	89.9%	#17 of 18	Official leaderboard
IOI v2 (Vals)	14.9%	#43 of 45	Official leaderboard
Vibe Code Bench 1-100	7.33%	#25 of 26	Official leaderboard
Code Migration v1.1	8.82%	#30 of 30	Official leaderboard
WeirdML v3	3.90%	#27 of 27	Official leaderboard
FrontierSWE	4.12%	#25 of 25	Official leaderboard
OWE-Bench	1.10%	#14 of 14	Official leaderboard

Agents and tools

Benchmark	Score	Rank	Source
BrowseComp	77.1%	#38 of 81	Lab model card
MCP Atlas	76%	#20 of 37	Official leaderboard
τ³-bench Banking	25%	#18 of 25	Official leaderboard
Chess Puzzles	21%	#67 of 93	Epoch AI run
Terminal-Bench 2.1 Extended	24.6%	#33 of 44	Official leaderboard
Terminal-Bench 2.1	63.8%	#43 of 56	Lab model card
RuneBench	3.76	#60 of 76	Official leaderboard
Toolathlon-Verified	45.5%	#22 of 25	Official leaderboard
BrowseComp Extended	4.50%	#39 of 42	Official leaderboard
Agent Arena	-17.32%	#54 of 56	Official leaderboard
Terminal-Bench 4.1	0%	#43 of 43	Official leaderboard
Terminal-Bench 4.0	0%	#40 of 40	Official leaderboard

Professional work

Benchmark	Score	Rank	Source
CorpFin v2	68.6%	#7 of 119	Official leaderboard
LEXam-hard	40.8%	#1 of 15	Official leaderboard
TaxEval v2	75.3%	#20 of 129	Official leaderboard
MedScribe	85.4%	#27 of 97	Official leaderboard
PolicyBench	87.7%	#15 of 50	Official leaderboard
LegalBench	83%	#64 of 138	Official leaderboard
MortgageTax	64%	#47 of 87	Official leaderboard
MedCode	41.2%	#53 of 96	Official leaderboard
Harvey LAB (Legal Agent Benchmark)	2.08%	#49 of 80	Official leaderboard
APEX-1	61.6%	#47 of 71	Official leaderboard
Legal Research Bench	28.4%	#62 of 93	Official leaderboard
Vals Finance Agent v2	46.6%	#54 of 80	Official leaderboard
TRLawBench	8.48	#23 of 34	Official leaderboard
Public Benefits Bench	58.5%	#36 of 50	Official leaderboard
Tax Agent Bench	15.3%	#63 of 84	Official leaderboard
Vals Index v2	34.1%	#49 of 65	Official leaderboard
MedXpertQA MM Extended	51%	#31 of 41	Official leaderboard
$OneMillion-Bench	22.1%	#10 of 13	Official leaderboard
GDPval-AA v2	1,238	#21 of 27	Lab model card
Excel Modeling Benchmark	38.4%	#61 of 76	Official leaderboard
JobBench	31.1%	#13 of 16	Official leaderboard
YC-Bench	$307,033	#50 of 59	Official leaderboard
Tuesday Work Index	32.2	#38 of 44	Official leaderboard
APEX-Agents	33.8%	#53 of 61	Collected by Epoch AI, source not named
APEX-Agents 1.1	48.8%	#57 of 65	Official leaderboard
DAYJOB: Healthcare	0.40%	#33 of 37	Official leaderboard
GDPval Extended	0.20%	#39 of 43	Official leaderboard
Harvey LAB v1.1 (HLAB-Vals-1.1)	3.75%	#35 of 38	Official leaderboard
Vals Finance Agent v2.1	49.9%	#38 of 40	Official leaderboard
GDPx	2.34%	#19 of 20	Official leaderboard
DeepFinance	13.1%	#22 of 23	Official leaderboard
APEX-Accounting	31.2%	#47 of 49	Official leaderboard
Vals Index v2.1	28.7%	#47 of 49	Official leaderboard
Vals Index v2.2	29%	#36 of 37	Official leaderboard
DAYJOB: Finance	0%	#38 of 39	Official leaderboard
Vals Finance Index	38.3%	#39 of 40	Official leaderboard
Excel Modeling Benchmark v1.1	41%	#41 of 42	Official leaderboard
Vals Legal Index	54.2%	#38 of 38	Official leaderboard
CAD Arena	0.12	#12 of 12	Official leaderboard

Vision, audio and robotics

Benchmark	Score	Rank	Source
Audio MC	56.6%	#3 of 18	Official leaderboard
VoiceBench	91.4%	#3 of 12	Lab model card
CharXiv	82%	#24 of 75	Lab model card
MMAU	77.2%	#6 of 15	Lab model card
MMMU-Pro	73.5%	#47 of 90	Lab model card
MMMU-Pro Extended	24%	#33 of 38	Official leaderboard
GlobeBench	11.9%	#17 of 19	Official leaderboard
CharXiv Extended	69.8%	#35 of 38	Official leaderboard
Chartography	10%	#41 of 43	Official leaderboard

Safety and security

Benchmark	Score	Rank	Source
StrongREJECT	98.6%	#1 of 19	Lab model card
BioTIER Permit	99.9%	#8 of 69	Official leaderboard
DistressBench	78.5%	#6 of 42	Official leaderboard
ABC Bench Fragment Design	95.7%	#7 of 27	Official leaderboard
nocot-bench (NCRI)	78.3	#62 of 185	Official leaderboard
Molecular Biology Capabilities Test	51.2%	#36 of 90	Official leaderboard
World-Class Bio	49.1%	#31 of 74	Official leaderboard
Agentic BAIM-LLM Evaluation (ABLE)	59.9%	#8 of 18	Official leaderboard
ABC Bench Liquid-Handling Robot	92%	#17 of 34	Official leaderboard
FORTRESS	25%	#38 of 73	Official leaderboard
Human Pathogen Capabilities Test	51.1%	#45 of 86	Official leaderboard
Human Pathogen Capabilities Test v2 (HPCT v2)	50.9%	#41 of 71	Official leaderboard
CyberBench v1.1 (Vals)	66.8%	#27 of 44	Official leaderboard
CyberBench (Vals)	64.1%	#17 of 26	Official leaderboard
BioTIER Refuse	30%	#48 of 69	Official leaderboard
ABC Bench Screening Evasion	55.9%	#12 of 17	Official leaderboard
VCT-v2	31.9%	#22 of 25	Official leaderboard
DeepsecBench	6.27	#27 of 29	Official leaderboard
SRE-Bench	0%	#34 of 35	Official leaderboard
CWE-bench	37%	#12 of 12	Official leaderboard

Other

Benchmark	Score	Rank	Source
ACCoRD	71.2	#46 of 152	Official leaderboard
Conceptual Reasoning Index	52.6	#61 of 151	Official leaderboard
LMCA	37.6%	#68 of 147	Official leaderboard
SAGE (Vals)	36.5%	#63 of 81	Official leaderboard
Vals Multimodal Index	49.4%	#27 of 33	Official leaderboard
LiveBench	44.3%	#29 of 31	Official leaderboard

Its lab's numbers put it at 150.1, 1.8 points above its ECI

The Self-Reported Index fits one score to the 16 capability results Thinking Machines reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 147.9 to 152.4. How we compute it.

Its lab's numbers next to 7 leaderboards

Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.

Benchmark	Lab says	Board says	Gap
SimpleQA Verified	43.9% SimpleQA Verified	40.3% Epoch AI run, xhigh	+3.6 pts
DesignArena Website	1,257 DesignArena, agentic web development, 1257 Elo	1,216 Leaderboard	+41
τ³-bench Banking	23.7% Tau 3 Banking (Artificial Analysis)	25% Leaderboard, max	−1.3 pts
GPQA Diamond	87.2% GPQA Diamond (Artificial Analysis)	88.3% Epoch AI run, xhigh	−1.1 pts
Audio MC	56.6% Audio MC	56.6% Leaderboard, thinking, text output	Same
MCP Atlas	76% MCP Atlas	76% Leaderboard, xhigh	Same
Toolathlon-Verified	45.5% Toolathlon Verified	45.5% Leaderboard, xhigh	Same