- GDPval (Elo) 1237.2
- τ³-bench Banking 30.2
- MultiChallenge 60.4
- AA-Omniscience-Public (Accuracy) 46.4
- IFBench 76.8
- GPQA Diamond 89.2
- AA-LCR v1.1 77.7
- MRCR v2 75.5
- SWE-Bench Atlas 51.1
- DeepSWE v1 46.3
- BrowseComp+ 55.2
- DeepSearchQA 71.3
Mercury 3
Mercury 3.
35
Benchmarks with a score
0
Ranked first, out of five or more models
0
In the top three
Inception Labs Oct 2026
Mercury 3
Closed weights Flagship
148.8
#104 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
- IFBench 76.8% #17 /80
- Legal Research Bench 28.9% #61 /93
- Vals Index v2.2 43.7% #33 /37
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 | 76.8% | #17 of 80 | Lab model card | |
| MRCR v2 | 75.5% | #13 of 30 | Lab model card | |
| MultiChallenge | 60.4% | #20 of 42 | Lab model card | |
| AA-LCR 1.1 | 77.7% | #4 of 6 | Lab model card |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GPQA Diamond | 89.2% | #41 of 302 | Lab model card |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| ALE-Bench | 1,692 | #45 of 126 | Official leaderboard | |
| DesignArena 3D Design | 1,224 | #49 of 136 | Official leaderboard | |
| DesignArena Game Dev | 1,205 | #62 of 143 | Official leaderboard | |
| DesignArena Website | 1,231 | #67 of 151 | Official leaderboard | |
| APEX-SWE | 43.1% | #29 of 65 | Official leaderboard | |
| DesignArena Code | 1,225 | #65 of 143 | Official leaderboard | |
| DesignArena UI Component | 1,226 | #64 of 139 | Official leaderboard | |
| Vibe Code Bench | 60.9% | #60 of 120 | Official leaderboard | |
| DesignArena SVG | 1,167 | #53 of 103 | Official leaderboard | |
| DesignArena Data Visualization | 1,207 | #63 of 118 | Official leaderboard | |
| SWE Atlas | 51.1% | #9 of 14 | Lab model card | |
| Code Migration v1.1 | 39.3% | #22 of 30 | Official leaderboard | |
| DeepSWE | 46.3% | #49 of 56 | Lab model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| LLM Chess | 1,308 | #12 of 123 | Official leaderboard | |
| τ³-bench Banking | 30.2% | #10 of 22 | Lab model card | |
| BrowseComp-Plus | 55.2% | #2 of 4 | Lab model card | |
| DeepSearchQA | 71.3% | #21 of 23 | Lab model card |
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| APEX-Agents 1.1 | 62.9% | #41 of 65 | Official leaderboard | |
| Legal Research Bench | 28.9% | #61 of 93 | Official leaderboard | |
| Tax Agent Bench | 18.2% | #58 of 84 | Official leaderboard | |
| Harvey LAB v1.1 (HLAB-Vals-1.1) | 5.00% | #30 of 38 | Official leaderboard | |
| Vals Legal Index | 63% | #32 of 38 | Official leaderboard | |
| APEX-Accounting | 50.8% | #42 of 49 | Official leaderboard | |
| Vals Finance Agent v2.1 | 54.7% | #35 of 40 | Official leaderboard | |
| Vals Index v2.2 | 43.7% | #33 of 37 | Official leaderboard | |
| Vals Finance Index | 47.2% | #36 of 40 | Official leaderboard | |
| Excel Modeling Benchmark v1.1 | 57.9% | #39 of 42 | Official leaderboard | |
| GDPval-AA v2.1 | 1,237 | #10 of 10 | Lab model card |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| nocot-bench (NCRI) | 72.3 | #88 of 185 | Official leaderboard | |
| OncologyBench | 8.00% | #26 of 30 | Official leaderboard |
Its lab's numbers put it at 148.8, number 104 of 404
The Self-Reported Index fits one score to the 8 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 145.8 to 151.7. How we compute it.
- GPQA Diamond 89.2
- IFBench 76.8
- MRCR v2 75.5
- DeepSearchQA 71.3
- MultiChallenge 60.4
- SWE Atlas 51.1
- DeepSWE 46.3
- τ³-bench Banking 30.2
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.