Cohere Embed v4 (Multimodal)

A dated reading of what is claimed, reported, and independently verified in the current evidence.

Current reading

The readiness gap, in one scan

AI-assisted assembly · derived results

Evidence strength Critical

Decision

Human editorial judgment · 2026-06-28

Adopt with guardrails

Why: GA on three major clouds, 128k context and native interleaved image+text embeddings are a genuine fit for multimodal commerce search and remain the only mainstream production multimodal option; main risks are image-token cost, sub-leader text accuracy, and vendor opacity on limits — all manageable with metering and own-data eval.
Next: Pilot on a single mixed-modality corpus (PDF + image product catalog), benchmark 1024-dim Matryoshka vs full 1536-dim for recall/cost tradeoff, validate multimodal recall on own data, and cap image-token spend with a budget alarm.

Constraints

No named blocker is present in the current public projection.

Evidence summary

AI-assisted assembly

Counts and dates only. Raw signals, private excerpts, trust records, and internal corpus material are not published here.

Publication record

Initial public reading

This is the initial public reading. No earlier readiness change is recorded.