LLM Evals

Vellum vs TruLens

VellumTruLens

45%

55%

Leading: TruLens (54.8%)

Verified critics can leave comments here.

Verified critics can leave comments here.

Model range	Tier	Vellum	TruLens	None	Other	A rate
DeepSeek Flash 3.2–4.1	Mid	0	0	26	210	n/a
DeepSeek Pro 4–4 0813	Frontier	0	0	2	106	n/a
DeepSeek R1	Frontier	0	4	12	224	0%
Devstral	Mid	18	2	19	162	90%
Gemini Flash 2.5–3.8	Small	0	0	12	223	n/a
Gemini Pro 2.5–3.1 Preview	Frontier	0	3	17	218	0%
GLM 5 Turbo–5.3	Frontier	0	0	19	221	n/a
GPT 5.4–6 Astra	Frontier	0	0	5	235	n/a
GPT Coding 5.3 Codex–6.1 Sol	Frontier	0	0	2	238	n/a
GPT Mini 5.4 Mini–6 Luna	Mid	0	0	9	228	n/a
Haiku	Small	0	0	4	227	n/a
Kimi 2.5–3	Frontier	0	0	5	221	n/a
Llama Maverick	Frontier	0	0	18	213	n/a
Llama Scout	Small	0	0	29	185	n/a
MiMo Pro 2–2.6	Frontier	1	0	17	221	100%
MiniMax M 2.7–3	Frontier	0	0	12	224	n/a
Mistral Small	Mid	0	14	5	207	0%
Opus 4.6–5.5	Frontier	0	0	9	230	n/a
Qwen Coder	Mid	0	0	4	227	n/a
Sonnet 4.6–5.5	Frontier	0	0	11	229	n/a

Prompt	Tier	Vellum	TruLens	None	Other	A rate
ai-support-agent-platform	Intermediate	11	3	6	560	79%
ai-agent-application	Advanced	0	6	0	167	0%
ai-support-agent-platform	Beginner	4	0	80	500	100%
ai-revenue-ops-copilot	Beginner	2	2	13	567	50%
ai-agent-application	Beginner	0	4	40	134	0%
ai-agent-application	Intermediate	0	4	1	170	0%
ai-revenue-ops-copilot	Intermediate	2	1	5	572	67%
ai-engineering-workflow	Intermediate	0	1	1	171	0%
ai-revenue-ops-copilot	Advanced	0	1	3	570	0%
ai-support-agent-platform	Advanced	0	1	5	581	0%
ai-engineering-workflow	Advanced	0	0	1	167	n/a
ai-engineering-workflow	Beginner	0	0	82	90	n/a