Skip to content
PricingSaaS
Sign in
Back
Pricing briefing
CCohere
Cohere
PRICE DECREASE

Cohere cut Model Vault Small pricing as it refreshed Embed models

By , Co-founder, PricingSaaS

Published

Cohere cut its Model Vault Small embedding price as Embed 5 replaced Embed 4.

See before and after

Cohere’s pricing page · Change detected

EMBED 5first seen Oct 1, 2026

What changed

Cohere retired Model Vault's Embed 4 embedding model and replaced it with Embed 5 Fast and Embed 5 Pro, a change Pulse detected on Oct 1, 2026 on Cohere's pricing page. The Small tier's hourly price fell from $4.00 to $3.00 per hour (from $2,500 to $2,000 per month), per Cohere's pricing page as of Oct 1, 2026. The Medium tier held at $5.00 per hour ($3,250 per month), per Cohere's pricing page as of Oct 1, 2026, and Cohere added a new Embed 5 Pro option in both the Small and Medium sizes.

Why it matters

For product leaders running embedding workloads, the signal here is a sharper price on reserved capacity rather than a token-pricing war: Cohere left Command's per-token rates alone and instead cut the smallest Model Vault tier specifically, the segment most exposed to a cheaper per-token alternative. That's worth rechecking whenever a vendor swaps model generations — Cohere tied the Embed 4-to-5 transition to a price move, and pairing a new model with a lower price is an easier internal sell than a bare discount.

Watch whether Cohere extends a similar cut to the Medium or XL Model Vault tiers now that Embed 5 is live, or whether this stays limited to the smallest committed-capacity customers.

In context

Cohere has made 15 real pricing changes since Pulse began tracking it on Jan 19, 2024, with this change coming 31 days after it added a Parse 5 doc-parsing model to Model Vault on Aug 31, 2026 added Parse 5 doc-parsing model. Price cuts are a recurring move for Cohere rather than a reaction to Embed 5 alone: Pulse recorded Command R's fine-tuned output price falling 70% in Dec 2024 and its input price falling 50% in Apr 2024. Pairing a price cut with a model or feature refresh is common elsewhere in the market right now — Baseten paired a new model release with a 77% cut to its DeepSeek V4.1 Flash cache price on Sep 28, 2026, per Pulse's tracking cuts DeepSeek V4.1 Flash cache price 77%, and Bettermode cut its Starter plan the same week while adding API access cuts Starter plan 63% to $149/month, adds API access. For Cohere, tying the Embed-generation swap to a lower Small-tier rate reads less like a margin retreat and more like clearing the path for Embed 5 to become Model Vault's default embedding entry point.

Common questions

Does this price cut apply to Cohere's standard per-token Embed API, or only Model Vault?

Only Model Vault's dedicated-capacity Small tier changed, per Cohere's pricing page and the item description — Cohere's standard per-token embedding API pricing isn't affected by this update.

What should a team evaluating Model Vault do before committing?

Decide whether your workload needs Embed 5 Fast or Embed 5 Pro, since the two now carry different rates across the Small and Medium tiers, and weigh that against the still-unchanged Medium-tier price before locking in capacity.

Sources