Preseason
MatchesRankingsPrompts
GitHub
Preseason
MatchesRankingsPromptsMethodologyContact

© 2026 Preseason. All rights reserved.

PrivacyTerms
@betocmn
AI / LLM Integration
Methodology

AI SDK vs Anthropic Claude API

AI SDKAIAI SDKvsANAnthropic Claude API
AI SDKAnthropic Claude API
34%
66%

Leading: Anthropic Claude API (65.5%)

Statistics

MetricValue
AI SDK wins40
Anthropic Claude API wins76
Abstains (no tool)128
Other tool chosen3744
Decisive cases116
AI SDK win rate (unweighted)34.5%
95% CI26.5% - 43.5%
AI SDK win rate (weighted)34.5%

Comments

AI SDK

No comments yet

Verified critics can leave comments here.

Anthropic Claude API

No comments yet

Verified critics can leave comments here.

Model range breakdown

Model rangeTierAI SDKAnthropic Claude APINoneOtherA rate
DeepSeek Flash 3.2–4.1Mid509194100%
DeepSeek Pro 4–4 0813Frontier00972n/a
DeepSeek R1Frontier0023187n/a
DevstralMid19016150100%
Gemini Flash 2.5–3.8Small001210n/a
Gemini Pro 2.5–3.1 PreviewFrontier002210n/a
GLM 5 Turbo–5.3Frontier1503193100%
GPT 5.4–6 AstraFrontier000207n/a
GPT Coding 5.3 Codex–6.1 SolFrontier000212n/a
GPT Mini 5.4 Mini–6 LunaMid001210n/a
HaikuSmall04901630%
Kimi 2.5–3Frontier003194n/a
Llama MaverickFrontier0027179n/a
Llama ScoutSmall0011192n/a
MiMo Pro 2–2.6Frontier004206n/a
MiniMax M 2.7–3Frontier11601866%
Mistral SmallMid0018156n/a
Opus 4.6–5.5Frontier000212n/a
Qwen CoderMid001209n/a
Sonnet 4.6–5.5Frontier01102020%

Per-prompt breakdown

PromptTierAI SDKAnthropic Claude APINoneOtherA rate
ai-support-agent-platformAdvanced438154610%
ai-support-agent-platformIntermediate166056973%
ai-revenue-ops-copilotAdvanced02245470%
ai-support-agent-platformBeginner183154586%
ai-revenue-ops-copilotIntermediate0515820%
ai-revenue-ops-copilotBeginner22155150%
ai-agent-applicationAdvanced0026149n/a
ai-agent-applicationBeginner0050128n/a
ai-agent-applicationIntermediate0044127n/a