Claude Haiku 5.5

Claude Haiku 5.5 is a model from Anthropic, released October 7, 2026. It ranks #17 of 36 models on the Vals Index with 58.73%. Its best result is #4 of 127 on Vibe Code Bench v1.1.

Release Date: Oct 7, 2026

Developer Anthropic ย ๐Ÿ‡บ๐Ÿ‡ธ
Context Window 1M
Max Output Tokens 128k
Token Costs (in/out) $0.10/0.50
Weights Private
Input Modalities

Accuracy

58.73 % ยฑ 1.28

Cost / Test (Vals Index)

$ 3.175

Latency

49 min 57 s

Fallback Rate 0.00%
Refusal Rate 0.22%
Vals Index
BenchmarksAccuracyRankings

58.73%

ยฑ1.28
17/36

57.45%

ยฑ1.80
19/39

70.29%

ยฑ1.83
17/37

38.50%

ยฑ1.67
13/24

52.36%

ยฑ3.49
14/30

75.48%

ยฑ5.21
6/48

65.48%

ยฑ3.69
33/41

65.37%

ยฑ2.06
11/39

61.79%

ยฑ1.64
23/26

43.75%

ยฑ3.45
28/93

51.54%

ยฑ2.17
15/107

89.54%

ยฑ1.92
6/109

35.13%

ยฑ3.21
11/30

89.00%

ยฑ3.15
12/53

89.83%

ยฑ1.44
3/29

63.19%

ยฑ1.25
25/50

5.72%

ยฑ1.44
17/37

31.92%

ยฑ3.35
28/84

20.41%

ยฑ3.79
11/26

90.44%

ยฑ1.73
4/127

70.37%

ยฑ1.33
14/31

4.17%

ยฑ1.52
33/37

100.00%

ยฑ0.00
3/25

47.28%

ยฑ10.72
30/46

2.50%

ยฑ1.11
9/66

43.96%

ยฑ2.27
10/22

36.36%

ยฑ2.31
22/63

10.00%

ยฑ3.61
19/45
Proprietary BenchmarksAcademic BenchmarksIndustry Partners
Vals
Default Provider : Anthropic
Temperature: Default
Top P: Default
Top K: Default
Max Output Tokens: 128,000
Compute Effort: max

Compared with

Updates

Oct 8, 2026

We evaluated Anthropicโ€™s new Claude Haiku 5.5 across our benchmark suite.

We ran Haiku 5.5 without a refusal fallback. Content-filter refusals are scored as failures, including 8 of 270 BioMysteryBench attempts and 1 of 70 Terminal-Bench-Science tasks.

Haiku 5.5 is priced at $0.10 per million input tokens and $0.50 per million output tokens. The model has a 1M-token context window and 128k max output tokens. Evaluations were run with compute effort set to โ€œmaxโ€.

Congrats to the Anthropic team on the release!