COMPARISON / EMBEDDINGS

Mistral Embed vs. Voyage 4 Lite vs. Cohere Embed 4

Models for semantic search, RAG and code retrieval. Pick two to four options to explore how they differ.

Snapshot: 2026-09-08 · Public sources · Not a real-time feed

Editorial edition · Reviewed on 2026-09-08 · Not a real-time feed

MiOption 1

Mistral AI · View details

VoOption 2

Voyage AI · View details

CoOption 3

Cohere · View details

Option 4

Same category, comparable metrics

AT A GLANCE

What changes between your options

CSV

Differences only Only metrics everyone has data forHighlighted = best value on that dimension, not an overall winner

Comparison of Mistral Embed, Voyage 4 Lite, Cohere Embed 4
Metric	Mistral Embed	Voyage 4 Lite	Cohere Embed 4
Text priceUSD / M tokens · lower is better	$0.10Vendor ↗	$0.02Vendor ↗	No dataNot included
Reference dimensionsdimensions · higher is better	1,024Vendor ↗	1,024Vendor ↗	1,536Vendor ↗
Input per documenttokens · higher is better	8KVendor ↗	32KVendor ↗	128KVendor ↗

Text pricing per million tokens, excluding starter credits and batch. Dimensions and context don't measure retrieval quality. Don't mix vector spaces from different models.

VISUAL COMPARISON

Compare each dimension

Metric

Text price

Embeddings API price for text; index storage is not included.

0$0.05$0.10

Voyage 4 Lite

$0.02

Mistral Embed

$0.10

1 with no data · left out of the chart

Price versus capacity

A relationship between two dimensions, not an automatic recommendation.

Select a product in the legend to isolate it. Points may overlap.

1 product lacks both values · not plotted

DETAILS THAT MATTER

Specs and operations

Feature	Mistral Embed	Voyage 4 Lite	Cohere Embed 4
Dimensiones de referencia	1024	1024 (por defecto)	1536 (por defecto)
Identificador API	mistral-embed	voyage-4-lite	embed-v4.0
Entrada	Texto	Texto; distinguir query y document	Texto, imagen y documentos mixtos
Dimensiones admitidas	No data	256, 512, 1024, 2048	256, 512, 1024, 1536

EVIDENCE, NOT ASSUMPTIONS

Sources, scope and fine print

What to validate before you make a decision.

Methodology

Mistral Embed

Modelo de texto; los vectores no son compatibles con los de otro modelo por tener la misma dimensión. Contexto publicado como 8k.

Checked on 2026-09-08 · Editorial update

Voyage 4 Lite

Tarifa posterior a créditos iniciales; no se descuentan créditos del comparador. Se muestran las 1024 dimensiones por defecto; también admite 256, 512 y 2048.

Checked on 2026-09-08 · Editorial update

Cohere Embed 4

No se incorpora una tarifa comparable por token: el precio depende de modalidad y despliegue. La oferta dedicada por instancia no se convierte artificialmente a tokens.

Checked on 2026-09-08 · Editorial update