Compare Models

Claude Sonnet 4.5 (Thinking)

Claude Sonnet 4.5 (Thinking) is a model from Anthropic, released September 29, 2025. Its best result is #39 of 151 on LegalBench.

Release Date: Sep 29, 2025

Developer Anthropic  🇺🇸
Context Window 1M
Max Output Tokens 64k
Token Costs (in/out) $3.00/15.00
Weights Private
Input Modalities

Accuracy

64.70 %

Avg. Cost (In/Out)

$ 3.00 / $ 15.00

Latency

6 min 13 s

Vals Index
BenchmarksAccuracyRankings
MedCode

44.13%

±2.00
42/107
MedScribe

84.10%

±1.87
39/109
MortgageTax

63.99%

±0.95
50/98
SAGE

36.06%

±3.21
67/91
TaxEval v2

73.30%

±0.86
51/145
Vibe Code Bench v1.1

22.62%

±3.73
79/115
GPQA Diamond

81.63%

±2.25
63/138
LiveCodeBench

73.00%

±1.13
83/143
LegalBench

84.08%

±0.45
39/151
MMLU Pro

87.36%

±0.39
28/138
MMMU Pro

79.31%

±0.97
48/93
SWE-bench

70.00%

±2.05
62/88
Terminal-Bench 2.0

41.57%

±5.25
29/67
Proprietary BenchmarksAcademic BenchmarksIndustry Partners
Vals
Default Provider : Anthropic
Temperature: 1
Top P: Default
Top K: Default
Max Output Tokens: 64,000

Updates

Sep 29, 2025

We ran the recently-released Claude Sonnet 4.5 (Thinking) our benchmarks, and found very strong performance:

  • On Finance Agent, it beats the previous state-of-the-art by five percentage points.
  • It also takes the #1 spot on SWE-bench Verified and Terminal-Bench, beating out GPT 5 Codex.
  • It is in the top 10 models on the majority of our benchmarks, and also showed better performance than Claude Sonnet 4 (Thinking) on almost all benchmarks.
  • It has a 1-million token context window, when the flag “context-1m-2025-08-07” is enabled

Overall, this model is extremely capable, at the same mid-range price point as its predecessor.