Latimal
Latimal Menu Intelligence

The world’s food, mapped by meaning.

The food model behind search and recommendations for delivery platforms. It places every dish by what it means, so a craving finds the right plate in any cuisine, in any language.

Get API keyTry the playground

5,000 free credits, no card. Concierge onboarding.

#1on FoodEval100+languagesp50 ~200msper query

Color is origin
Benchmarks

How Latimal compares

Built for food and benchmarked against the strongest general-purpose models.

Overall

Category averages across the three tables below.

  • All tasks from FoodEval, the public food-embedding benchmark.
  • Search: production API. Matching and classification: raw embeddings + cosine.
  • Average: unweighted mean of the three category scores.
Overall: Latimal compared with seven embedding systems. Higher is better.
TaskLatimalFood Embed v1OpenAItext-embedding-3-largeVoyage AIVoyage 4 LargeCohereEmbed v4AlibabaGTE-largeNomic AINomic v1.5BAAIBGE-M3MicrosoftE5-large
Average3 categories0.8190.6820.6590.6430.6140.6240.6090.505
SearchProduction NDCG@10, 4 tasks0.8690.4550.4470.4510.4270.4220.4080.411
MatchingMean F1, 7 tasks0.8510.7580.7410.7410.6990.7390.7180.704
ClassificationMacro F1, 1 task0.7380.8330.7890.7370.7160.7100.7010.399

Matching

Best F1, 7 tasks.

  • Same-dish detection across cuisines, scripts, and noise levels.
  • Raw embeddings + cosine, no reranking.
Matching: Latimal compared with seven embedding systems. Higher is better.
TaskLatimalFood Embed v1OpenAItext-embedding-3-largeVoyage AIVoyage 4 LargeCohereEmbed v4AlibabaGTE-largeNomic AINomic v1.5BAAIBGE-M3MicrosoftE5-large
Average7 tasks0.8510.7580.7410.7410.6990.7390.7180.704
Indian cuisine0.8170.7450.7180.7320.7050.7310.7110.680
Global cuisine0.8670.8280.7830.8290.6950.7320.7160.716
Beverages0.7460.7150.7190.7100.7100.7150.7060.706
Bakery & desserts0.7550.7350.7150.6910.6820.6840.6840.688
Portion size0.9720.8490.7910.8350.7250.8550.8210.757
Noisy menu0.9160.6850.6400.6670.6720.7500.6740.648
Cross-lingual0.8860.7480.8200.7210.7070.7070.7170.731

Search

Production search, NDCG@10.

  • Latimal: production API, measured at the public API boundary. Reproducible with an API key.
  • Competitors: embedding + bge-reranker-v2-m3.
Search: Latimal compared with seven embedding systems. Higher is better.
TaskLatimalOpenAItext-embedding-3-largeVoyage AIVoyage 4 LargeCohereEmbed v4AlibabaGTE-largeNomic AINomic v1.5BAAIBGE-M3MicrosoftE5-large
Average4 tasks0.8690.4550.4470.4510.4270.4220.4080.411
Food searchNDCG@100.9380.5900.5900.5890.5720.5640.5520.554
Concept searchNDCG@100.8090.4050.3920.3910.3740.3570.3360.328
Diet & allergen searchNDCG@100.8020.1720.1610.1650.1350.1320.1320.136
Noisy searchNDCG@100.9250.6530.6440.6600.6280.6350.6140.628

Diet & allergen search: 4.7x the best competitor.

Classification

Macro F1, 1 task. Linear probe on frozen embeddings, 26 menu classes.

Classification: Latimal compared with seven embedding systems. Higher is better.
TaskLatimalFood Embed v1OpenAItext-embedding-3-largeVoyage AIVoyage 4 LargeCohereEmbed v4AlibabaGTE-largeNomic AINomic v1.5BAAIBGE-M3MicrosoftE5-large
Cuisine classificationMacro F10.7380.8330.7890.7370.7160.7100.7010.399

All models compared at 384 dimensions. June 2026. Full benchmarks on Hugging Face →

Food Embed v1 ranks #1 of 10 on the FoodEval leaderboard. FoodEval leaderboard →

One API, the whole menu

Everything a delivery platform needs to understand food.

Diners feel search and recommendations directly. Dedup, classification, and health scoring keep the catalog clean behind the scenes.

Semantic search

A diner types “something brothy and warming” and the right bowls rise to the top, whatever they are called on the menu.

RamenPhoLaksaDan Dan Mian

Smart recommendations

A cart of Ramen and Gyoza suggests what truly pairs. A craving for “cold & refreshing” answers across cuisines.

EdamameGreen TeaMochi
Iced MatchaHorchataCeviche

Cross-lingual matching

Recognize a dish across languages and scripts. シャワルマ, shawarma and شاورما resolve to one item.

Menu deduplication

Collapse Murgh Makhani, Butter Chicken and बटर चिकन into one canonical entry across restaurants.

Cuisine classification

Sort any item into its cuisine, from Levantine to Andean, including the ones general models miss.

It also cleans menu noise, flags dietary conflicts, and grades catalog health.
Built for aggregator search, cart intelligence, and menu operations.
Integration

Up and running in three calls.

One REST API. Nothing to host.

01Get your key
02Send a call
03Get results
POST /search

{ "query": "dumplings",
  "corpus": ["Gyoza", "Pierogi", "Gazpacho", "Empanada"] }

// => Gyoza (0.95), Pierogi (0.93), Empanada (0.91)
Get started

Start free. Pay only for what you run.

Every account gets 5,000 credits to try the whole API on a real menu before paying anything. Then you buy credits as you go.

Concierge onboarding for every account. Tell us what you are building and we'll set you up with a key.
5,000credits free
  • Credits never expire. Buy once, spend at your pace.
  • A search costs 0.05 credits per item, so 5,000 credits covers a full pilot menu.
  • No credit card, no seats, no monthly minimum.
Get API key

Private by default

Your menus never train our models.

Built for scale

p50 ~200ms per query, plus bulk endpoints.

Production-ready

Warm-failover standby, 99.5% uptime.

No infra to host

No GPUs, no model ops, no version pinning.

Put the whole world’s food on the map.

Give your platform search and recommendations that speak every language. Start with 5,000 free credits.

Get API keyRead the docs