Mercury 3

Inception’s hosted text-diffusion LLM launched October 19. Compare its native API contract, advertised 1M context and temporary launch discount.

Model Specificationsinception

Hosted Inception API; no local hardware requirement established.

Standard $0.20 input / $0.70 output per 1M tokens; temporary launch $0.10 / $0.35 with unspecified expiry. Cached input: 90% off input price.

Scoped Terms

🚀Function Overview

Mercury 3 generates text through diffusion. Evaluate latency and task accuracy on your workload; publisher benchmarks are not universal guarantees.

Native model ID: mercury-3; OpenAI-compatible chat completions endpoint.
Advertised context: 1M tokens. The quick-start output budget of 8,192 is an example setting, not a documented maximum.
Standard USD per 1M tokens: $0.20 input and $0.70 output. Launch promotion: $0.10 input and $0.35 output; the announcement gives no expiry. Cached tokens receive a 90% input-price discount.
Hosted access is announced through Inception, OpenRouter, Baseten and Vercel; verify each provider’s own limits and prices.
No inference request was executed here. Publisher speed and benchmark results are configuration-specific claims.

string

mercury-3 on the native Inception endpoint.

array

Role/content chat messages as documented by the native quick start.

string

The quick start uses medium; verify the full parameter contract before changing it.

Hardware Type: Hosted Inception API; no local hardware requirement established.
Commercial Use: Hosted commercial service under Inception terms; no downloadable-weight license is established.
Pricing: Standard $0.20 input / $0.70 output per 1M tokens; temporary launch $0.10 / $0.35 with unspecified expiry. Cached input: 90% off input price.