Mercury 2
Inception · released Mar 4, 2026
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...
Specification
- Context window
- 128K
- Max output
- 50K
- Knowledge cutoff
- Not stated
- Parameters
- Undisclosed
- Licence
- Proprietary
- Serving providers
- 1
- Moderated
- No
- Uptime
- 100.0%
Intelligence
21.9
63th percentile
Coding
31.1
Coding Index
Agentic
9.5
Agentic Index
Output speed
228 t/s
Median across providers
Latency
572ms
Time to first token
Cost per task
$0.07
Estimated
Benchmarks
Where the score comes from
The Intelligence Index is a composite. These are the underlying evaluations this model was actually measured on.
Evaluation scores
Percentage correct · higher is better
- GPQA Diamond77.0%
- τ²-bench (Telecom)70.8%
- IFBench69.8%
- AA-LCR (long context)40.7%
- SciCode38.7%
- Terminal-Bench Hard26.5%
- Humanity's Last Exam17.1%
An evaluation missing from this list was not run for this model — it is not a zero.
View as table
| Evaluation | Score |
|---|---|
| GPQA Diamond | 77.0% |
| τ²-bench (Telecom) | 70.8% |
| IFBench | 69.8% |
| AA-LCR (long context) | 40.7% |
| SciCode | 38.7% |
| Terminal-Bench Hard | 26.5% |
| Humanity's Last Exam | 17.1% |
Against its peers
Intelligence Index · this model highlighted, nearest peers in grey
- GLM 4.6Z.ai23.4
- GLM 4.7 FlashZ.ai23.3
- Mercury 2Inception21.9
- Qwen3.5-9BQwen21.8
- V3.1 TerminusDeepSeek21.7
- Qwen3 Coder NextQwen21.3
- R1DeepSeek20.4
- North Mini Code (free)Cohere20.2
Peers are the models sitting closest on the Intelligence Index — the set you would realistically choose between.
View as table
| Model | Intelligence |
|---|---|
| GLM 4.6 | 23.4 |
| GLM 4.7 Flash | 23.3 |
| Mercury 2 | 21.9 |
| Qwen3.5-9B | 21.8 |
| V3.1 Terminus | 21.7 |
| Qwen3 Coder Next | 21.3 |
| R1 | 20.4 |
| North Mini Code (free) | 20.2 |
Percentile among all indexed models
Pricing
What it costs to run
List prices per million tokens, plus what one representative task works out to.
List price
- Input / 1M tokens
- $0.25
- Output / 1M tokens
- $0.75
- Cached input / 1M
- $0.025
- Blended 3:1
- $0.375
One task, estimated
$0.07
- Input tokens
- 50,000
- Output tokens
- 80,000
- Profile
- Reasoning
Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).
Serving providers
Speed and latency figures are medians across these providers, so a widely-served model reports a blend rather than any single endpoint.
Arena
Head-to-head generation quality
Elo from pairwise judgements, broken out by the kind of thing the model was asked to build.
- Overall Elo
- 1,038
- Win rate
- 23.6%
- Strongest at
- 3D scenes
- Tournaments
- 2,089
Elo by category
Dot position on a 975–1,050 scale · Elo has no meaningful zero
- 1,038
- 1,017
- 1,016
- 1,016
- 1,011
- 992
Agent categories (full-stack apps, mobile apps) are only scored for models tested in agent harnesses.
View as table
| Category | Elo | Win rate |
|---|---|---|
| 3D scenes | 1038 | 23.6% |
| Websites | 1017 | 19.9% |
| SVG | 1016 | 24.1% |
| Game development | 1016 | 20.6% |
| Data visualisation | 1011 | 21.4% |
| UI components | 992 | 18.0% |