Mercury 3
Inception · v3 · depuis 19. Oktober 2026 · 18× · vu le 21 oct. 2026
—
Momentum
Mercury 3 is the flagship language model from Inception, a diffusion-based LLM (dLLM) that generates tokens in parallel rather than sequentially. Inception released it on October 19, 2026 and describes it as its most capable diffusion LLM to date. According to the vendor, it generates 1,068 tokens per second on NVIDIA GPUs; independent quality benchmarks are not yet available.
Historique du momentum
23.07.21.10.
Fonctionnalités
| Context Window (Tokens) | 1,000,000 tokens (max output 131,072) |
| Multimodality | Text only (input and output) |
| Platform | Inception API, OpenRouter, Baseten, Vercel; OpenAI-compatible API |
| Price per 1M Tokens | List $0.20 input / $0.70 output; launch price $0.10 / $0.35 (50% off); cached input $0.02 |
| Release Date | October 19, 2026 |