Preseason
MatchesRankingsPrompts
GitHub
Preseason
MatchesRankingsPromptsMethodologyContact

© 2026 Preseason. All rights reserved.

Privacy PolicyTerms & Conditions
LLM Evals
Methodology

Vellum vs Arize Phoenix

VEVellumvsArize PhoenixARArize Phoenix
VellumArize Phoenix
48%
52%

Leading: Arize Phoenix (52.5%)

Statistics

MetricValue
Vellum wins19
Arize Phoenix wins21
Abstains (no tool)90
Other tool chosen2314
Decisive cases40
Vellum win rate (unweighted)47.5%
95% CI32.9% - 62.5%
Vellum win rate (weighted)47.5%

Comments

Vellum

No comments yet

Verified critics can leave comments here.

Arize Phoenix

No comments yet

Verified critics can leave comments here.

Per-model breakdown

ModelTierVellumArize PhoenixNoneOtherA rate
Devstral 2 2512Mid1804103100%
Qwen3 Coder NextMid0831200%
MiniMax M2.7Frontier0551190%
GPT 5.4 MiniMid0431250%
Llama 4 ScoutSmall0241150%
Mistral Small 4Mid0211210%
MiMo V2 ProFrontier108123100%
Claude Haiku 4.5Small001124n/a
Claude Opus 4.6Frontier000132n/a
Claude Sonnet 4.6Frontier000132n/a
DeepSeek R1 0528Frontier007125n/a
DeepSeek V3.2Mid0022106n/a
Gemini 2.5 FlashSmall001126n/a
Gemini 2.5 ProFrontier009123n/a
GLM 5 TurboFrontier0019113n/a
GPT 5.3 CodexFrontier000132n/a
GPT 5.4Frontier000132n/a
Kimi K2.5Frontier003116n/a
Llama 4 MaverickFrontier000127n/a

Per-prompt breakdown

PromptTierVellumArize PhoenixNoneOtherA rate
ai-support-agent-platformIntermediate111539292%
ai-support-agent-platformBeginner476433636%
ai-revenue-ops-copilotBeginner261039225%
ai-revenue-ops-copilotIntermediate23439540%
ai-revenue-ops-copilotAdvanced0423940%
ai-support-agent-platformAdvanced005405n/a