Anthropic: Claude Haiku 5.5

Anthropic

budget

Capabilities

VisionToolsStructured OutputExtended ThinkingLong Context

Strengths

Exceptional cost-efficiency for budget workloads
Strong vision and multimodal capabilities for tier
1M context window enables long-document tasks
Fast enough for real-time subagent orchestration

Weaknesses

Slowest speed score (0.12) limits latency-sensitive applications
Lower reasoning and analysis depth vs. flagship models
Closed weights prevent full data control via self-hosting
Limited code generation precision vs. specialist models

Pricing

Input / 1M tokens

$0.09999999999999999

Output / 1M tokens

$0.5

Context window

1,000k

Transparency

Open weights

0.0 / 10

Open training data

0.0 / 10

Open methodology

3.5 / 10

Licence openness

1.0 / 10

Provider disclosure

5.0 / 10

FMTI company score

N/A

Composite:4.0 / 10

Anthropic publishes constitutional AI methods and safety research, but model weights and training data remain closed. Limited public disclosure of training details compared to peers.

Sustainability

Inference energy

N/A

Training footprint

N/A

Provider infrastructure

5.5 / 10

Composite:5.0 / 10

Small model size suggests lower per-inference energy cost, but provider sustainability commitments not explicitly disclosed. Anthropic's infrastructure choices remain largely opaque.

Task Fitness

question answering

64%

code

65%

mathematics

50%

business communication

50%

vision

70%

analysis

64%

extraction

72%

long-form generation

68%

research

50%

reasoning

64%

summarisation

75%

translation

70%

conversation

72%