Inception model product
Mercury 2 is the fastest reasoning LLM, built on diffusion-based language model (dLLM) architecture.
Updated Aug 12, 2026. Default version: Mercury 2
Technical details for the model's default version.
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Benchmark scores for Mercury 2.
| IFBench | 71.0% | 15 | 29 | 50.0% | C | |
| SciCode | 38.0% | 16 | 19 | 16.7% | C | |
| Tau2 Airline | 53.0% | 19 | 23 | 18.2% | C | |
| LiveCodeBench | 67.0% | 24 | 73 | 68.1% | C | |
| AIME 2025 | 91.1% | 43 | 114 | 62.8% | C | |
| GPQA | 74.0% | 114 | 234 | 51.5% | C |
Preference and agent-evaluation results for the default version.
| webdev | overall | 108 | 1166.2 | 908 | N/A | |
| text factuality | overall | 145 | 1371.9 | 3,126 | N/A | |
| text | overall | 173 | 1357.3 | 3,127 | N/A | |
| text style control | overall | 198 | 1346.4 | 3,127 | N/A |
Provider-specific output speed and catalog latency for Mercury 2. Runtime does not affect the capability score.
| Inception | 437.018 tok/s | 2.958 s | 128K | 8.2K |
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
Official vendor API pricing appears first, followed by individual provider offers.
| NanoGPT | mercury-2 | global | $0.25 | $0.75 | 128K | |
| Inception | mercury-2 | global | $0.25 | $0.75 | 128K | |
| Vercel AI Gateway | inception/mercury-2 | global | $0.25 | $0.75 | 128K | |
| OpenRouter | inception/mercury-2 | global | $0.25 | $0.75 | 128K | |
| Venice AI | mercury-2 | global | $0.3125 | $0.9375 | 128K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Available versions of this model. The score column identifies the version used in the overall ranking.
| Mercury 2 | 39.3 | N/A | 128K | 8.2K | No | Proprietary |
Key information about Mercury 2 and its available data.
Mercury 2 is the fastest reasoning LLM, built on diffusion-based language model (dLLM) architecture. Instead of generating text token-by-token, it refines multiple text blocks simultaneously, achieving over 1,000 tokens per second on Nvidia Blackwell GPUs — 5x faster than leading speed-optimized LLMs.
Supports tool usage and JSON output with 128K context window.
Data as of 2026-08-11.
Open a comparison with the three ranked models immediately above and below this model.
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
Common questions about Mercury 2.
Mercury 2's default version was released on Feb 24, 2026.
Mercury 2's official API price is $0.25 per million input tokens and $0.75 per million output tokens via Inception. The lowest tracked third-party offer starts at $0.25 input and $0.75 output via NanoGPT.
Mercury 2 was created by Inception.
The default version has a 128K token context window.
No. The default version is not marked as having publicly available weights.
5 provider offerings are linked to the default version.
Nearby ranked alternatives include Qwen3 VL 32B Thinking, DeepSeek-V3.1, Claude Sonnet 4.