Compare models
Pick two to four models. No markup on model usage: you pay the provider's list price.
| Attribute |   DeepSeek V4 Pro |  Muse Spark 1.3 |
|---|---|---|
| Provider | DeepSeek | Meta |
| Model id | deepseek/deepseek-v4-pro | meta/muse-spark-1.3 |
| Status | Available via Safeguard | Available via Safeguard |
| Input / 1M tokens | $1.32 | $1.25 |
| Output / 1M tokens | $3.96 | $4.25 |
| Price notes | Peak-hour list price (cache miss). Off-peak is half: $0.66 in / $1.98 out. Cache-hit input $0.044 peak. DeepSeek says it is phasing V4 Pro out but continues to serve it. | Standard tier (no training on your data). Cached input $0.15. No long-context premium. Contributor tier (muse-spark-1.3-contributor) is $0.10 / $0.20 in exchange for allowing training. Audio input not fully supported on 1.3. |
| Context | 1M | 1.05M |
| Parameters | Not published | Not published |
| Max output | 384,000 tokens | Not published |
| Modality | Text | Text and image |
| Weights | Open | Closed |
| Zero data retention | Not stated | Not stated |
| Prompts go to | DeepSeek (People's Republic of China) | Meta (Meta Model API) |
| Retention | Privacy policy: personal data stored in the PRC and kept as long as necessary to provide the service, may be used to train models, with a right to opt out. No API-specific retention period or ZDR offering published | Standard tier: prompts and completions not used to train Meta models, Contributor tier allows training in exchange for lower prices. No retention period published in the API docs |
| Region | China (PRC) | Not stated |
| Price checked | 2026-10-08 | 2026-10-08 |
Self-healing security runs on Safeguard.
Your first fix PR is minutes away.
No sales call required, even your agent can complete the purchase over MCP.