Introducing Mercury 2.5 (Inception blog) Sep 8, 2026
Mercury 2.5
Mercury 2.5 ranks first on tau3-bench.
26
Benchmarks with a score
1
Ranked first, out of five or more models
1
In the top three
Inception Labs Sep 2026
Mercury 2.5
Closed weights
141.8
#165 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
- IFBench 77% #16 /80
- SciCode 38% #35 /49
- AA-LCR (Artificial Analysis Long Context Reasoning) 68% #13 /38
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| IFBench | 77% | #16 of 80 | Lab model card | |
| AA-LCR (Artificial Analysis Long Context Reasoning) | 68% | #13 of 38 | Lab model card | |
| StudyArena Text | 1,238 | #36 of 85 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GPQA Diamond | 79% | #125 of 302 | Lab model card | |
| ProofBench v1.1 | 3.00% | #50 of 52 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SciCode | 38% | #35 of 49 | Lab model card | |
| ALE-Bench | 977 | #113 of 126 | Official leaderboard | |
| Vibe Code Bench | 3.68% | #109 of 120 | Official leaderboard | |
| Code Migration | 4.52% | #74 of 78 | Official leaderboard | |
| IOI v2 (Vals) | 2.44% | #45 of 45 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| tau3-bench | 96% | #1 of 14 | Lab model card | |
| DeepSearchQA | 34% | #23 of 23 | Lab model card |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LegalBench | 83.1% | #61 of 138 | Official leaderboard | |
| CorpBench Work | 75.7 | #11 of 17 | Official leaderboard | |
| MedCode | 31.3% | #88 of 96 | Official leaderboard | |
| Public Benefits Bench | 45.3% | #47 of 50 | Official leaderboard | |
| Legal Research Bench | 4.33% | #88 of 93 | Official leaderboard | |
| Vals Finance Agent v2 | 18.9% | #77 of 80 | Official leaderboard | |
| MedScribe | 55.1% | #94 of 97 | Official leaderboard | |
| Vals Index v2 | 12.9% | #64 of 65 | Official leaderboard | |
| Excel Modeling Benchmark | 8.79% | #75 of 76 | Official leaderboard | |
| Tax Agent Bench | 2.49% | #84 of 84 | Official leaderboard | |
| Harvey LAB (Legal Agent Benchmark) | 0% | #80 of 80 | Official leaderboard | |
| Vals Index v2.1 | 8.86% | #49 of 49 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| nocot-bench (NCRI) | 65.8 | #117 of 185 | Official leaderboard | |
| CyberBench v1.1 (Vals) | 45.5% | #37 of 44 | Official leaderboard |
Its lab's numbers put it at 141.8, number 165 of 404
The Self-Reported Index fits one score to the 6 capability results Inception Labs reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 133.1 to 150.5. How we compute it.
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.