Forkelo ranking · Updated October 2026
7 cheesecakes across San Francisco, ranked on what people say about the dish itself — never the restaurant's overall rating.
Noe Valley Bakery · Noe Valley
11 dish-specific mentions · 100% positive · 1 source type · menu verified October 2026
Light, fluffy individual cheesecake crowned with fresh raspberries that regulars call a favorite. Seasonal pumpkin is a hit too.
16 dish-specific mentions · 81% positive · 1 source type · menu verified October 2026
Creamy, not too sweet Basque cheesecakes in many flavors that fans seek out. Some found slices small for the price or one too burned.
12 dish-specific mentions · 75% positive · 1 source type · menu verified October 2026
Rich, creamy New York style slices, plain or berry, that fans call among the best they have had. A few found them not dense enough.
Vive La Tarte · Noe Valley
14 dish-specific mentions · 86% positive · 1 source type · menu verified October 2026
A light, creamy, not too sweet cheesecake that many fans call the best in SF. One recent whole cake came out goopy and underset.
b. patisserie · Lower Pacific Heights
11 dish-specific mentions · 73% positive · 1 source type · menu verified October 2026
Creamy, lightly sweet seasonal cheesecakes like blueberry lime and pumpkin that many favor. A few found them overpriced or just fine.
Schubert's Bakery · Inner Richmond
11 dish-specific mentions · 64% positive · 1 source type · menu verified October 2026
Light fruit cheesecakes, strawberry and mango above all, that longtime fans love. Others found them too light or store bought tasting.
Uncle Tetsu · Stonestown
9 dish-specific mentions · 56% positive · 1 source type · menu verified October 2026
Soft, fluffy Japanese cheesecake that fans love warm. Some found it closer to angel food cake, dry late in the day or damaged.
Every score is built from evidence about the specific dish: how often locals recommend it, how strongly, how recently, and whether independent sources agree. A restaurant's overall star rating is never converted into a dish score.
Quality and confidence are reported separately. A dish two people loved can score high, but it is marked limited data until more evidence backs it up. The whole calculation — every constant, and three dishes worked through line by line — is written out on the methodology page.
Think we got it wrong? Tell us what’s better.
