MedicalBenchmark
Anthropic: Claude 3.5 Haiku provider

Claude 3.5 Haiku

235

#235 of 319 models in the general ranking

Cumulative performance across 3 MIR exams

Net score

432.66 pts

Accuracy

78.7%

Correct / Incorrect

472 / 118

Total Cost

$1.20

Overall Performance

(vs. average)
Accuracy

78.7%

avg: 80.6%

Net score

432.66 pts

avg: 453.30 pts

Correct

472

avg: 483

Incorrect

118

avg: 90

Total Cost

$1.20

avg: $9.58

Average response time

7.0s

avg: 17.9s

Output Tokens

227K

avg: 1.3M

Reasoning Tokens

0

avg: 898K

Average confidence

97.9%

avg: 95.4%

Breakdown by Exam

MIR 2024
231
Correct
160
Incorrect
37
Accuracy
80.0%
Net score
147.66
MIR 2025
240
Correct
148
Incorrect
49
Accuracy
74.0%
Net score
131.66
MIR 2026
228
Correct
164
Incorrect
32
Accuracy
82.0%
Net score
153.33