Claude Opus 4.1 vs Claude Sonnet 4
Wins 5 of 7 areas
Coding · Agents · Reasoning · Images and charts · Long documents
Wins 1 of 7 areas
Facts
Claude Opus 4.1 is the stronger all-rounder.Claude Sonnet 4 is cheaper and better at facts.
Scores updated · 26 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.
- CodingWriting and fixing software20Claude Opus 4.12 of 3 tests · 1 tie
- ReasoningHard problems that need careful thinking20Claude Opus 4.12 of 3 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own10Claude Opus 4.11 of 1 test
- Images and chartsUnderstanding pictures, charts and video10Claude Opus 4.11 of 1 test
- Long documentsFinding answers in very long texts10Claude Opus 4.11 of 1 test
- FactsGetting facts right instead of making them up01Claude Sonnet 41 of 1 test
- Following instructionsDoing exactly what it is asked00Even0 each · 2 ties
Agents, facts, images and charts and long documents 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 80% 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 Opus 4.1 pulls ahead
- Harder college exam questions with imagesMMMU-Pro+6.1points ahead
- Reasons across sets of long documentsAA-LCR+5.7points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+3.2points ahead
Where Claude Sonnet 4 pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 26 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Opus 4.1
- Terminal-Bench HardClaude Opus 4.1 by 3.234.331.1+3.2
- SWE-bench VerifiedClaude Opus 4.1 by 1.874.572.7+1.8
- SciCodetie40.940tie
ReasoningClaude Opus 4.1
- GPQA DiamondClaude Opus 4.1 by 3.280.977.7+3.2
- Humanity's Last ExamClaude Opus 4.1 by 1.812.510.7+1.8
- CritPttie00.3tie
FactsClaude Sonnet 4
- Vectara HHEM hallucination ratelower is betterClaude Sonnet 4 by 1.511.810.3+1.5
Images and chartsClaude Opus 4.1
- MMMU-ProClaude Opus 4.1 by 6.167.961.8+6.1
Following instructionsEven
- IFBenchtie55.454.7tie
- Multi-Challengetie57.257.1tie
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.
- τ²-BenchClaude Opus 4.1 by 26.371.545.2+26.3
- τ²-Bench (Retail)Claude Opus 4.1 by 11.886.875+11.8
- Terminal-BenchClaude Opus 4.1 by 7.843.335.5+7.8
- AIME 2025Claude Opus 4.1 by 7.57870.5+7.5
- τ²-Bench Telecom (AA run)Claude Opus 4.1 by 6.871.464.6+6.8
- vectara_answer_rateClaude Sonnet 4 by 6.292.498.6+6.2
- frontiermath_tier_4_v1Claude Opus 4.1 by 4.24.20+4.2
- TAU-bench (airline)Claude Sonnet 4 by 45660+4
- AA IntelligenceClaude Opus 4.1 by 3.922.818.9+3.9
- MMMLUClaude Opus 4.1 by 389.586.5+3
- TAU-bench (retail)Claude Opus 4.1 by 1.982.480.5+1.9
- vectara_avg_summary_lengthClaude Sonnet 4 by 16.7 rating points129.1145.8+16.7 rating
- vectara_factual_consistencyClaude Sonnet 4 by 1.588.289.7+1.5
- Artificial Analysis Coding IndexClaude Sonnet 4 by 1.136.537.6+1.1
Questions people ask
Which is better, Claude Opus 4.1 or Claude Sonnet 4?
Claude Opus 4.1 wins five of the seven areas where both have results: coding, agents, reasoning, images and charts and long documents. Claude Sonnet 4 wins facts, and costs 80% less. They are level on following instructions.
Which is better for coding?
Claude Opus 4.1. It wins 2 of the 3 coding tests both models report; Claude Sonnet 4 wins none, and 1 is a tie.
Which is cheaper?
Claude Opus 4.1 costs $15.00 per million input tokens and $75.00 per million output tokens; Claude Sonnet 4 costs $3.00 and $15.00. That makes Claude Sonnet 4 about 80% cheaper for the same work.
How do you compare the two?
We use the 26 benchmark tests both models have published scores on. The verdict counts the 12 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.