Claude 4 Sonnet vs Sarvam 105B

Side-by-side benchmark comparison across agentic, coding, multimodal, knowledge, reasoning, and math workflows.

Agentic
Coding
Multimodal & Grounded
Reasoning
Knowledge
Instruction Following
Multilingual
Mathematics

Claude 4 Sonnet· Sarvam 105B

Quick Verdict

Pick Claude 4 Sonnet if you want the stronger benchmark profile. Sarvam 105B only becomes the better choice if knowledge is the priority or you want the stronger reasoning-first profile.

Claude 4 Sonnet has the cleaner overall profile here, landing at 62 versus 60. It is a real lead, but still close enough that category-level strengths matter more than the headline number.

Claude 4 Sonnet's sharpest advantage is in agentic, where it averages 57.9 against 49.5. The single biggest benchmark swing on the page is SWE-bench Verified, 72.7% to 45%. Sarvam 105B does hit back in knowledge, so the answer changes if that is the part of the workload you care about most.

Sarvam 105B is the reasoning model in the pair, while Claude 4 Sonnet 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. Claude 4 Sonnet gives you the larger context window at 200K, compared with 128K for Sarvam 105B.

Operational tradeoffs

PricePricing unavailableFree*
Speed40 t/sN/A
TTFT1.33sN/A
Context200K128K

Decision framing

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.

BenchmarkClaude 4 SonnetSarvam 105B
AgenticClaude 4 Sonnet wins
Terminal-Bench 2.056%
BrowseComp62%49.5%
OSWorld-Verified57%
CodingClaude 4 Sonnet wins
HumanEval65%
SWE-bench Verified72.7%45%
LiveCodeBench36%
SWE-bench Pro48%
LiveCodeBench v671.7%
Multimodal & Grounded
MMMU-Pro81%
OfficeQA Pro78%
Reasoning
MuSR69%
BBH82%
LongBench v271%
MRCRv272%
gpqaDiamond78.7%
hle11.2%
KnowledgeSarvam 105B wins
MMLU86.5%90.6%
GPQA73%
SuperGPQA71%
MMLU-Pro75%81.7%
HLE12%
FrontierScience67%
SimpleQA71%
Instruction FollowingSarvam 105B wins
IFEval83%84.8%
Multilingual
MGSM84%
MMLU-ProX81%
MathematicsSarvam 105B wins
AIME 202373%
AIME 202475%
AIME 202574%88.3%
HMMT Feb 202369%
HMMT Feb 202471%
HMMT Feb 202570%
BRUMO 202572%
MATH-50081%98.6%
HMMT Feb 202585.8%
HMMT Nov 202585.8%
Frequently Asked Questions (6)

Which is better, Claude 4 Sonnet or Sarvam 105B?

Claude 4 Sonnet is ahead overall, 62 to 60. The biggest single separator in this matchup is SWE-bench Verified, where the scores are 72.7% and 45%.

Which is better for knowledge tasks, Claude 4 Sonnet or Sarvam 105B?

Sarvam 105B has the edge for knowledge tasks in this comparison, averaging 81.7 versus 57.8. Inside this category, MMLU-Pro is the benchmark that creates the most daylight between them.

Which is better for coding, Claude 4 Sonnet or Sarvam 105B?

Claude 4 Sonnet has the edge for coding in this comparison, averaging 49.1 versus 45. Inside this category, SWE-bench Verified is the benchmark that creates the most daylight between them.

Which is better for math, Claude 4 Sonnet or Sarvam 105B?

Sarvam 105B has the edge for math in this comparison, averaging 92.3 versus 75.1. Inside this category, MATH-500 is the benchmark that creates the most daylight between them.

Which is better for agentic tasks, Claude 4 Sonnet or Sarvam 105B?

Claude 4 Sonnet has the edge for agentic tasks in this comparison, averaging 57.9 versus 49.5. Inside this category, BrowseComp is the benchmark that creates the most daylight between them.

Which is better for instruction following, Claude 4 Sonnet or Sarvam 105B?

Sarvam 105B has the edge for instruction following in this comparison, averaging 84.8 versus 83. Inside this category, IFEval is the benchmark that creates the most daylight between them.

Last updated: April 3, 2026

Weekly LLM Benchmark Digest

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.