Relace is a San Francisco startup that builds fast models and inference infrastructure for coding agents. Its products include Relace Apply, for merging code edits, and Relace Search for retrieval across large repositories. It is backed by Y Combinator and Andreessen Horowitz.
- Founded
- 2022
- Type
- Private company
- Founders
- Preston Zhou and Eitan Borgnia
- AI lab
- Inference provider
- 13 hosted endpoints
| Model | Endpoint | Context | Input /M | Output /M | Cached input /M | Cache write /M | Uptime 1d |
|---|---|---|---|---|---|---|---|
 DeepSeek V4 Flash 0423 | relace/fp4 | 1M | $0.05 | $0.45 | $0.05 | — | 100.00% |
| DeepSeek V4 Flash 0731 | relace/fp4 | 1M | $0.07 | $0.14 | $0.014 | — | 99.93% |
| DeepSeek V4 Pro 0423 | relace/fp4 | 1M | $0.90 | $1.50 | $0.25 | — | 100.00% |
| DeepSeek V4 Pro 0813 | relace/fp4 | 1M | $0.0999 | $8.00 | $0.0999 | — | 100.00% |
| DeepSeek V4.1 Flash | relace/fp4 | 1M | $0.12 | $0.50 | $0.003 | — | 100.00% |
| DeepSeek V4.1 Flash 1017 | relace/fp4 | 1M | $0.12 | $0.50 | $0.003 | — | 100.00% |
 GLM 5.2 | relace/fp4 | 1M | $0.0066 | $7.9833 | $0.0066 | — | 99.31% |
| GLM 5.3 | relace/fp4 | 1M | $0.0703 | $4.3843 | $0.0703 | — | 100.00% |
| GLM 5.3 Flash | relace/fp4 | 1M | $0.0399 | $0.0784 | $0.03 | — | 100.00% |
 Kimi K3 | relace/fp4 | 1M | $0.1824 | $11.0055 | $0.18 | — | 100.00% |
| Kimi K3.1 Pro | relace/fp4 | 1M | $0.2375 | $8.4085 | $0.20 | — | 100.00% |
 Relace Apply 3 | relace/fp8 | 256K | $0.85 | $1.25 | — | — | — |
| Relace Search | relace/bf16 | 256K | $1.00 | $3.00 | — | — | 100.00% |