Inception: Mercury 2

● Excellent

by Inception Fast Reasoning Rank #52 of 75

8.1
/ 10

Mercury 2 is an extremely fast reasoning LLM, and the first reasoning diffusion LLM (dLLM). Instead of generating tokens sequentially, Mercury 2 produces and refines multiple tokens in parallel, achieving...

Choose a model to compare against Inception: Mercury 2

Specifications

Specifications for Inception: Mercury 2
AttributeValue
Lab Inception
Tags Fast Reasoning
Overall Score 8.1/10
Release Date 2026-03
Context Window 128,000 tokens
Input Price / 1M $0.25
Output Price / 1M $0.75
Input Modalities Text
Output Modalities Text

Strengths

  • First reasoning diffusion LLM — parallel token generation
  • Extremely fast inference via parallel refinement
  • Novel architecture for reasoning tasks
  • Competitive pricing

Weaknesses

  • Experimental architecture with unknown edge cases
  • Limited track record
  • Narrower general knowledge vs traditional LLMs

Best For

  • Latency-critical reasoning
  • High-throughput classification
  • Experimental LLM workflows

Sources & Further Reading

Related Models

Scores are aggregated from public benchmarks (MMLU, HumanEval, MATH, GSM8K, LMSYS) and normalized to a 1–10 scale. Methodology →