Claude Sonnet 4 vs Grok 3
Wins 5 of 7 areas
Coding · Agents · Reasoning · Long documents · Following instructions
Wins 2 of 7 areas
Facts · Images and charts
Claude Sonnet 4 is the stronger all-rounder.Grok 3 is better at facts.
Scores updated · 28 tests both models report · How we compare
Where each one wins
Tests won in each of the seven areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- ReasoningHard problems that need careful thinking40Claude Sonnet 44 of 5 tests · 1 tie
- CodingWriting and fixing software20Claude Sonnet 42 of 2 tests
- AgentsCarrying out multi-step tasks on its own10Claude Sonnet 41 of 1 test
- Long documentsFinding answers in very long texts10Claude Sonnet 41 of 1 test
- Following instructionsDoing exactly what it is asked10Claude Sonnet 41 of 1 test
- FactsGetting facts right instead of making them up12Grok 32 of 3 tests
- Images and chartsUnderstanding pictures, charts and video01Grok 31 of 1 test
Agents, images and charts, long documents and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
Claude Sonnet 4 costs 25% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Claude Sonnet 4 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+58.2points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+19.7points ahead
- Reasons across sets of long documentsAA-LCR+12.3points ahead
Where Grok 3 pulls ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+6.2points ahead
- College exam questions with charts, maps and diagramsMMMU+3.6points ahead
Every test, side by side
All 28 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Sonnet 4
- Terminal-Bench HardClaude Sonnet 4 by 19.731.111.4+19.7
- SciCodeClaude Sonnet 4 by 3.24036.8+3.2
ReasoningClaude Sonnet 4
- SimpleBenchClaude Sonnet 4 by 9.445.536.1+9.4
- GPQA DiamondClaude Sonnet 4 by 8.477.769.3+8.4
- Humanity's Last ExamClaude Sonnet 4 by 6.610.74.1+6.6
- ARC-AGI-2Claude Sonnet 4 by 5.95.90+5.9
- CritPttie0.30tie
FactsGrok 3
- AA-Omniscience · Non-hallucinationClaude Sonnet 4 by 58.270.912.7+58.2
- AA-Omniscience · AccuracyGrok 3 by 6.222.728.9+6.2
- Vectara HHEM hallucination ratelower is betterGrok 3 by 4.510.35.8+4.5
Following instructionsClaude Sonnet 4
- IFBenchClaude Sonnet 4 by 7.854.746.9+7.8
Other results14 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- SimpleQAGrok 3 by 66.115.982+66.1
- ARC-AGI-1Claude Sonnet 4 by 34.5405.5+34.5
- AA-OmniscienceClaude Sonnet 4 by 33.40.2-33.2+33.4
- AIME 2025Grok 3 by 22.870.593.3+22.8
- BrowseComp-ZHClaude Sonnet 4 by 18.329.110.8+18.3
- Artificial Analysis Coding IndexClaude Sonnet 4 by 17.837.619.8+17.8
- τ²-Bench Telecom (AA run)Claude Sonnet 4 by 15.864.648.8+15.8
- LiveCodeBenchGrok 3 by 10.968.579.4+10.9
- AA Agentic IndexGrok 3 by 717.624.6+7
- AA IntelligenceClaude Sonnet 4 by 6.818.912.1+6.8
- vectara_answer_rateClaude Sonnet 4 by 5.698.693+5.6
- vectara_avg_summary_lengthClaude Sonnet 4 by 49.9 rating points145.895.9+49.9 rating
- vectara_factual_consistencyGrok 3 by 4.589.794.2+4.5
- frontiermath_tier_4_v1tie00tie
Questions people ask
Which is better, Claude Sonnet 4 or Grok 3?
Claude Sonnet 4 wins five of the seven areas where both have results: coding, agents, reasoning, long documents and following instructions. Grok 3 wins facts and images and charts.
Which is better for coding?
Claude Sonnet 4. It wins 2 of the 2 coding tests both models report; Grok 3 wins none.
Which is cheaper?
Claude Sonnet 4 costs $3.00 per million input tokens and $15.00 per million output tokens; Grok 3 costs $4.00 and $20.00. That makes Claude Sonnet 4 about 25% cheaper for the same work.
How do you compare the two?
We use the 28 benchmark tests both models have published scores on. The verdict counts the 14 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 14 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.