AI BENCHY
Advertise here
#254

Mercury 2.5 Preview

Inception Release: 2026-09-02 Tested on: 2026-09-02 02:04 inception/mercury-2.5-preview::low
~100BOtherClosedEstimated
(high) (medium) (low) (none)

Summary

Mercury 2.5 Preview scores 5.0 on AI BENCHY and ranks #254. It has 10.0 reliability, a 47.0% pass rate, $0.011 total cost, and 1.31s average response time.

What makes Mercury 2.5 Preview unique: It stands out most in Puzzle Solving, where it ranks #1, while Tool Calling is its weakest area at #15. It is notably fast compared with similar models.

Model facts

Researched on 2026-09-02

Estimated
Parameters
~100B
Architecture
Other
Availability
Closed
License
-

Best estimate from public evidence; the vendor did not disclose every value. Inception identifies Mercury 2.5 as a diffusion LLM but does not disclose its parameter count; the estimate follows the Mercury 2 family record.

Parameter source Availability source

Score

5.0

Consistency

7.0

Reliability

10.0

Total Cost (Current Price)

$0.011

Total Output Tokens

31,975

Total Input Tokens

133,525

Input Price

$0.040 / 1M

Output Price

$0.150 / 1M

Tests Correct

Wrong Tests: 16

Attempt pass rate: 47.0%

Flaky tests

8

Flaky tests had mixed outcomes across runs (at least one pass and one fail).

Response Time (avg)

1.31s

Response Time (max): 7.29s

Response Time (total): 28.77s

Wrong answer: 12 Did not follow instructions: 2 API error: 1 No answer: 1

Hamster playing table tennis

Prompt: Create a detailed SVG illustration of a hamster playing table tennis.

#254 Mercury 2.5 Preview

low
Cost
$0.001
Time
1.5s
Tokens
1,169 tok

Charts

Choose the first model, then click a second model to open a side-by-side page.

Total Output Tokens

Score vs Total Output Tokens

Quick Compare

Mercury 2.5 PreviewlowvsMiniMax M2.7mediumMercury 2.5 PreviewlowvsLaguna S 2.1lowFree AvailableMercury 2.5 PreviewlowvsNorth Mini CodenoneFree AvailableMercury 2.5 PreviewlowvsDeepSeek V3.2noneMercury 2.5 PreviewlowvsQwen3.5-9BnoneMercury 2.5 PreviewlowvsGPT-4o-mininoneMercury 2.5 PreviewlowvsMiMo-V2.5noneMercury 2.5 PreviewlowvsMercury 2.5 PreviewnoneMercury 2.5 PreviewlowvsQwen3 Coder NextnoneMercury 2.5 PreviewlowvsLing-2.6-flashnone

Category Breakdown

Category Score Consistency Tests Correct
Anti-AI Tricks 5.6 3.8
Coding 5.5 10.0
Combined 3.0 10.0
Data parsing and extraction 6.3 5.8
Domain specific 4.1 4.4
General Intelligence 5.0 10.0
Instructions following 6.3 10.0
Puzzle Solving 8.2 7.2
Tool Calling 2.8 1.6
Trivia 3.0 10.0

Compared models