Compare Models

Claude Haiku 4.5 (Thinking)

Claude Haiku 4.5 (Thinking) is a model from Anthropic, released October 15, 2025. Its best result is #32 of 109 on MedScribe.

Release Date: Oct 15, 2025

Developer Anthropic  🇺🇸
Context Window 200k
Max Output Tokens 64k
Token Costs (in/out) $1.00/5.00
Weights Private
Input Modalities

Accuracy

42.35 %

Avg. Cost (In/Out)

$ 1.00 / $ 5.00

Latency

4 min 28 s

Vals Index
BenchmarksAccuracyRankings
Code Migration

7.55%

±1.06
67/79
EMB

22.66%

±2.39
70/76
Finance Agent (v2)

31.01%

±0.56
72/81
Legal Research Bench

10.58%

±2.14
79/90
MedCode

32.68%

±2.00
88/107
MedScribe

85.23%

±1.90
32/109
MortgageTax

62.16%

±0.96
56/98
Tax Agent Bench

5.86%

±1.54
76/81
TaxEval v2

67.54%

±0.91
108/145
Vibe Code Bench v1.1

11.39%

±3.13
105/124
GPQA Diamond

72.22%

±2.47
90/138
LiveCodeBench

41.17%

±1.11
130/143
LegalBench

81.24%

±0.49
83/151
MMLU Pro

78.72%

±0.48
102/138
MMMU Pro

46.07%

±1.20
91/93
ProgramBench

0.00%

±0.00
57/64
SWE-bench

66.60%

±2.11
71/88
Terminal-Bench 2.1

43.82%

±1.95
65/75
Proprietary BenchmarksAcademic BenchmarksIndustry Partners
Vals
Default Provider : Anthropic
Temperature: Default
Top P: Default
Top K: Default
Max Output Tokens: 64,000

Updates

Oct 16, 2025

We evaluated Claude Haiku 4.5 (Thinking) and found strong performance:

  • The model places 3rd on our Vals Index, demonstrating well-rounded capabilities across diverse tasks.
  • On Terminal-Bench, Haiku 4.5 achieves 3rd place, showing particular strength on coding tasks.
  • While it performs well on certain coding benchmarks, the model achieves middle-of-the-pack performance on most other benchmarks.
  • The model struggles significantly on our proprietary CaseLaw benchmark and the public MedQA, GPQA Diamond, MMLU Pro, MMMU Pro, and LiveCodeBench benchmarks.
  • Compared to Claude Sonnet 4.5 (Thinking), Haiku trades some performance for significantly faster speed and lower cost - see our model comparison for details.

Overall, Haiku 4.5 sits firmly on the Pareto frontier.