- Math500 97
- Humaneval 92.1
- MBPP 92.7
- Live Code Bench v6 70
- MMLU 85.1
- MMLU Pro 80
- Arena Hard v2 49
- GPQA Diamond 66.5
- AIME 25 (80.0 without tools, 96.7 with tools) 80
- HMMT Feb 2025 73.3
- HMMT Nov 2025 74.2
- Beyond AIME 58.3
- BrowseComp 35.5
- SWE-Bench Verified 34
- Tau2 (avg.) 45.7
- Average pairwise win rate across all dimensions on the Indic benchmark (indivibe) 89
Sarvam 30B
Sarvam 30B ranks first on MBPP.
16
Benchmarks with a score
1
Ranked first, out of five or more models
1
In the top three
Sarvam AI Mar 2026
Sarvam 30B
Open weights Small
132.7
#254 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
- BrowseComp 35.5% #74 /81
- tau2-bench 45.7% #75 /79
- MMLU-Pro 80% #80 /167
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 | |
|---|---|---|---|---|
| MMLU | 85.1% | #55 of 131 | Lab model card | |
| MMLU-Pro | 80% | #80 of 167 | Lab model card | |
| Arena-Hard v2.0 | 49% | #21 of 24 | Lab model card | |
| IndiVibe | 89% | #2 of 2 | Lab model card |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MATH-500 | 97% | #21 of 70 | Lab model card | |
| AIME 2025 | 80% | #89 of 143 | Lab model card | |
| GPQA Diamond | 66.5% | #197 of 302 | Lab model card | |
| Beyond AIME | 58.3% | #10 of 14 | Lab model card | |
| HMMT 2025 | 73.3% | #45 of 54 | Lab model card | |
| HMMT Nov 2025 | 74.2% | #27 of 28 | Lab model card |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MBPP | 92.7% | #1 of 71 | Lab model card | |
| HumanEval | 92.1% | #8 of 101 | Lab model card | |
| LiveCodeBench v6 | 70% | #56 of 90 | Lab model card | |
| SWE-bench Verified | 34% | #150 of 155 | Lab model card |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| BrowseComp | 35.5% | #74 of 81 | Lab model card | |
| tau2-bench | 45.7% | #75 of 79 | Lab model card |
Its lab's numbers put it at 132.7, number 254 of 404
The Self-Reported Index fits one score to the 15 capability results Sarvam AI 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 130.6 to 134.8. How we compute it.
- MATH-500 97
- MBPP 92.7
- HumanEval 92.1
- MMLU 85.1
- MMLU-Pro 80
- AIME 2025 80
- HMMT Nov 2025 74.2
- HMMT 2025 73.3
- LiveCodeBench v6 70
- GPQA Diamond 66.5
- Beyond AIME 58.3
- Arena-Hard v2.0 49
- tau2-bench 45.7
- BrowseComp 35.5
- SWE-bench Verified 34
Its launch post
The benchmarks the lab chose to headline, with the scores it gave.