Compare Models

Inkling

Inkling is an open-weight model from Thinking Machines, released July 15, 2026. It ranks #46 of 48 models on the Vals Index with 28.67%. Its best result is #28 of 109 on MedScribe.

Release Date: Jul 15, 2026

Developer Thinking Machines  🇺🇸
Context Window 256k
Max Output Tokens 256k
Token Costs (in/out) $1.00/4.05
Weights Open
Input Modalities

Accuracy

28.67 % ± 1.10

Cost / Test (Vals Index)

$ 1.463

Latency

20 min 7 s

Vals Index
BenchmarksAccuracyRankings
Vals Index

28.67%

±1.10
46/48
Code Migration

11.79%

±3.40
62/77
CyberBench v1.1

66.79%

±5.95
30/47
EMB

38.36%

±3.27
59/74
Finance Agent (v2)

46.60%

±0.93
52/78
Legal Research Bench

28.36%

±3.13
49/78
MedCode

41.19%

±2.23
56/107
MedScribe

85.41%

±1.84
28/109
MortgageTax

63.99%

±0.94
51/98
ProofBench v1.1

0.00%

±0.00
49/50
SAGE

36.55%

±3.22
66/91
Public Benefits Bench v1.1

58.46%

±1.28
36/50
SRE Bench

0.00%

±0.00
35/35
Tax Agent Bench

15.29%

±2.42
50/69
TaxEval v2

75.31%

±0.84
20/145
Vibe Code Bench 1-100

7.33%

±1.84
24/25
Vibe Code Bench v1.1

19.21%

±3.66
83/112
Harvey's Legal Agent Benchmark

2.08%

±0.83
48/79
GPQA Diamond

87.12%

±1.78
40/138
IOI

14.94%

±5.40
42/44
LiveCodeBench

85.52%

±1.00
32/143
LegalBench

82.95%

±0.50
64/151
MMLU Pro

86.30%

±0.34
43/138
MMMU Pro

78.50%

±0.99
51/93
ProgramBench

0.00%

±0.00
61/64
SkillsBench

26.09%

±3.86
34/35
SWE-bench

77.60%

±1.87
34/88
Terminal-Bench 4.0

0.51%

±0.51
45/48
Proprietary BenchmarksAcademic BenchmarksIndustry Partners
Vals
Default Provider : Thinkingmachines
Temperature: 1
Top P: 1
Top K: Default
Max Output Tokens: 256,000
Reasoning Effort: 0.99

Compared with

Inkling vs GPT 5.4 Nano 18 shared benchmarks Inkling vs Qwen 3.6 Plus 19 shared benchmarks Inkling vs Gemini 3.5 Flash Lite 21 shared benchmarks Inkling vs Qwen 3.7 Plus 10 shared benchmarks Inkling vs GPT 5.4 Mini 18 shared benchmarks Inkling vs MiMo V2.5 19 shared benchmarks Inkling vs MiMo V2.5 Pro 18 shared benchmarks Inkling vs Gemini 3.1 Pro Preview (02/26) 26 shared benchmarks

Updates

Jul 15, 2026

We evaluated Thinking Machines’ first open-weights model, Inkling, on the Vals Index.

  • Inkling scores 34.10% on the Vals Index, placing #30 of 43 models overall and #9 among open-weight models.

  • Its strongest component results include 75.49% on the SWE-bench Verified Vals Index subset and 69.23% on the CorpFin v2 Vals Index subset. It also scores 45.97% on the Finance Agent v2 Index subset.

  • Inkling scores 13.28% on the Vibe Code Bench Vals Index subset and 47.57% across three full trials of Terminal-Bench 2.1.

We used Thinking Machines’ recommended evaluation settings: temperature 1, top-p 1, up to 256k output tokens, separate reasoning enabled, and reasoning effort set to "0.99". Inkling supports a 256k context window, image inputs, and tool calling.

Congrats to the Thinking Machines team on the release!