Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.
Phi-4
40
3/8 categoriesSarvam 30B
48
Winner · 1/8 categoriesPhi-4· Sarvam 30B
Pick Sarvam 30B if you want the stronger benchmark profile. Phi-4 only becomes the better choice if coding is the priority or you would rather avoid the extra latency and token burn of a reasoning model.
Sarvam 30B is clearly ahead on the aggregate, 48 to 40. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
Sarvam 30B's sharpest advantage is in knowledge, where it averages 80 against 53.6. The single biggest benchmark swing on the page is HumanEval, 82.6% to 92.1%. Phi-4 does hit back in coding, so the answer changes if that is the part of the workload you care about most.
Sarvam 30B is the reasoning model in the pair, while Phi-4 is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use. Sarvam 30B gives you the larger context window at 64K, compared with 16K for Phi-4.
BenchLM keeps the benchmark table and the operator tradeoffs on the same page so a better score does not hide a materially slower, pricier, or smaller-context model.
Runtime metrics show N/A when BenchLM does not have a sourced snapshot for that exact model. The scoring rules and freshness policy are documented on the methodology page.
| Benchmark | Phi-4 | Sarvam 30B |
|---|---|---|
| AgenticPhi-4 wins | ||
| Terminal-Bench 2.0 | 44% | — |
| BrowseComp | 35% | 35.5% |
| OSWorld-Verified | 34% | — |
| CodingPhi-4 wins | ||
| HumanEval | 82.6% | 92.1% |
| SWE-bench Pro | 55% | — |
| LiveCodeBench v6 | — | 70.0% |
| SWE-bench Verified | — | 34% |
| Multimodal & Grounded | ||
| MMMU-Pro | 54% | — |
| OfficeQA Pro | 38% | — |
| Reasoning | ||
| LongBench v2 | 30% | — |
| MRCRv2 | 33% | — |
| gpqaDiamond | — | 66.5% |
| KnowledgeSarvam 30B wins | ||
| MMLU | 84.8% | 85.1% |
| GPQA | 56.1% | — |
| FrontierScience | 52% | — |
| MMLU-Pro | — | 80% |
| Instruction Following | ||
| Coming soon | ||
| Multilingual | ||
| MGSM | 80.6% | — |
| MMLU-ProX | 60% | — |
| MathematicsPhi-4 wins | ||
| MATH-500 | 94.6% | 97% |
| AIME 2025 | — | 80% |
| HMMT Feb 2025 | — | 73.3% |
| HMMT Nov 2025 | — | 74.2% |
Sarvam 30B is ahead overall, 48 to 40. The biggest single separator in this matchup is HumanEval, where the scores are 82.6% and 92.1%.
Sarvam 30B has the edge for knowledge tasks in this comparison, averaging 80 versus 53.6. Inside this category, MMLU is the benchmark that creates the most daylight between them.
Phi-4 has the edge for coding in this comparison, averaging 55 versus 34. Inside this category, HumanEval is the benchmark that creates the most daylight between them.
Phi-4 has the edge for math in this comparison, averaging 94.6 versus 86.5. Inside this category, MATH-500 is the benchmark that creates the most daylight between them.
Phi-4 has the edge for agentic tasks in this comparison, averaging 38.3 versus 35.5. Inside this category, BrowseComp is the benchmark that creates the most daylight between them.
Get notified when new models drop, benchmark scores change, or the leaderboard shifts. One email per week.
Free. No spam. Unsubscribe anytime. We only store derived location metadata for consent routing.