Very long inputs
Choose Embed v4.0
Embed v4.0 reads up to 128k tokens per request, against 8k for Gemini Embedding 2.
Choose Embed v4.0 for very long inputs. Choose Gemini Embedding 2 for the latest release.
Each answer below comes straight from the published prices, specifications and independent benchmark results.
Choose Embed v4.0
Embed v4.0 reads up to 128k tokens per request, against 8k for Gemini Embedding 2.
Choose Gemini Embedding 2
Gemini Embedding 2 is the newer release (April 2026, against April 2025 for Embed v4.0).
| Embed v4.0 | Gemini Embedding 2 | |
|---|---|---|
| Provider | Cohere | |
| Category | Embeddings | Embeddings |
| Status | Legacy | Generally available |
| Headline price | No data | $0.20 / 1M tokens |
| Context window | 128k | 8k |
| Max output | No data | No data |
| Released | April 2025 | April 2026 |
| Knowledge cutoff | No data | No data |
| Licence | Proprietary | Proprietary |
| Capabilities | Vision | Vision, Audio input, Video input |
| API model id | embed-v4.0 | gemini-embedding-2 |
Choose Embed v4.0 for very long inputs. Choose Gemini Embedding 2 for the latest release.
A price comparison is not possible: one of them has no comparable published price.
Embed v4.0 reads more per request: 128k tokens against 8k.
Embed v4.0 is made by Cohere and Gemini Embedding 2 by Google. A price comparison is not possible: one of them has no comparable published price. Embed v4.0 reads more per request: 128k tokens against 8k. Gemini Embedding 2 is the newer model (April 2026, against April 2025). Only Gemini Embedding 2 offers audio input, video input.
Gemini Embedding 2 was released on April 2026 and Embed v4.0 on April 2025.