Humanity's Last PRD: the AI benchmark for product managers

GPT-6 Astra leads by 2 points.

Combined score by model and task type. Choose a column heading to sort by that task.
Model		Product Craft						Growth				Metrics & Experimentation				Strategy						Leadership
Model
GPT-6 AstraOpenAI	89#1	96Best	81	84Best	94Best	70	93	91	85	98Best	100Best	76	96Best	89	99Best	88Best	82	91	92	98Best	84	95Best	84	90Best
GPT-6.1 SolOpenAI	87#2	94	83	72	93	89	94	88	87	84	96	73	78	93	99	82	90Best	89	95Best	81	80	85	96Best	87
Sonnet 5.5Anthropic	87#3	80	87Best	83	92	93Best	96Best	93Best	92Best	80	89	91Best	91	94Best	96	83	77	80	83	86	81	82	89	76
Haiku 5.5Anthropic	84#4	77	78	67	86	92	93	88	71	80	100	89	76	83	95	84	90	92Best	79	83	74	84	75	87
Opus 5.5Anthropic	82#5	74	82	68	93	77	91	79	73	90	93	78	88	81	83	74	60	84	95	82	85Best	80	84	83
GPT-6 LunaOpenAI	80#6	85	78	67	84	85	79	88	68	88	98	73	70	90	78	77	89	83	86	79	70	69	81	79
Gemini 4 ArgonGoogle	69#7	67	70	54	77	89	71	68	67	69	76	66	64	70	71	58	77	56	57	72	52	65	75	89
Gemini 3.8 FlashGoogle	61#8	62	57	39	60	76	54	59	65	76	73	56	52	70	59	58	50	44	60	65	48	61	69	79
Gemini 3.5 Flash-LiteGoogle	50#9	53	42	40	45	49	55	47	44	66	76	32	53	44	50	52	45	42	38	58	40	61	60	56

Combined scoreEach cell is the model’s best setup on that task type.

Tested with criteria sourced from the very best.

We carefully research and design the rubric for each task around the advice of guests on Lenny’s Podcast.

30 product leaders behind the checks

We credit their ideas. None of them has reviewed or endorsed the benchmark.

Show

Haiku 5.5 · API scores within 5 points of GPT-6 Astra · ChatGPT at about a twenty-fifth of the cost.

Best valueNobody beats these on both score and costCost partly estimated (app runs with no usage meter)At least one critical failure