Claude Sonnet 4.5 vs Mercury 2
Wins 5 of 6 areas
Coding · Agents · Reasoning · Facts · Long documents
Wins 1 of 6 areas
Following instructions
Claude Sonnet 4.5 is the stronger all-rounder.Mercury 2 is cheaper and better at following instructions.
Scores updated · 25 tests both models report · How we compare
Where each one wins
Tests won in each of the six 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 software40Claude Sonnet 4.54 of 4 tests
- AgentsCarrying out multi-step tasks on its own20Claude Sonnet 4.52 of 3 tests · 1 tie
- FactsGetting facts right instead of making them up20Claude Sonnet 4.52 of 3 tests · 1 tie
- ReasoningHard problems that need careful thinking10Claude Sonnet 4.51 of 3 tests · 2 ties
- Long documentsFinding answers in very long texts10Claude Sonnet 4.51 of 1 test
- Following instructionsDoing exactly what it is asked01Mercury 21 of 1 test
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.
Mercury 2 costs 94% 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 4.5 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+42.1points ahead
- Reasons across sets of long documentsAA-LCR+28.6points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+28.5points ahead
Where Mercury 2 pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+12.5points ahead
Every test, side by side
All 25 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Sonnet 4.5
- Terminal-Bench 2.1Claude Sonnet 4.5 by 28.555.827.3+28.5
- LMArena · WebDevClaude Sonnet 4.5 by 228 rating points13941166+228 rating
- Terminal-Bench HardClaude Sonnet 4.5 by 9.135.626.5+9.1
- SciCodeClaude Sonnet 4.5 by 845.737.7+8
AgentsClaude Sonnet 4.5
- GDPValClaude Sonnet 4.5 by 20.520.50+20.5
- τ-Bench V3 · BankingClaude Sonnet 4.5 by 1524.59.5+15
- Terminal-Bench 4.0tie00tie
ReasoningClaude Sonnet 4.5
- GPQA DiamondClaude Sonnet 4.5 by 6.483.477+6.4
- Humanity's Last Examtie17.817.1tie
- CritPttie1.10.8tie
FactsClaude Sonnet 4.5
- AA-Omniscience · Non-hallucinationClaude Sonnet 4.5 by 42.150.98.8+42.1
- AA-Omniscience · AccuracyClaude Sonnet 4.5 by 11.732.921.2+11.7
- Vectara HHEM hallucination ratelower is bettertie1212.3tie
Long documentsClaude Sonnet 4.5
- AA-LCRClaude Sonnet 4.5 by 28.672.343.7+28.6
Following instructionsMercury 2
- IFBenchMercury 2 by 12.557.369.8+12.5
Other results10 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.
- AA-OmniscienceClaude Sonnet 4.5 by 50.6-0.1-50.7+50.6
- Artificial Analysis Coding IndexClaude Sonnet 4.5 by 2152.131.1+21
- AA Agentic IndexClaude Sonnet 4.5 by 13.517.54+13.5
- τ²-Bench Telecom (AA run)Claude Sonnet 4.5 by 7.378.170.8+7.3
- AA IntelligenceClaude Sonnet 4.5 by 6.920.713.8+6.9
- vectara_answer_rateMercury 2 by 4.495.6100+4.4
- AIME 2025Mercury 2 by 4.18791.1+4.1
- LiveCodeBenchClaude Sonnet 4.5 by 47167+4
- vectara_avg_summary_lengthMercury 2 by 21.3 rating points127.8149.1+21.3 rating
- vectara_factual_consistencytie8887.7tie
Questions people ask
Which is better, Claude Sonnet 4.5 or Mercury 2?
Claude Sonnet 4.5 wins five of the six areas where both have results: coding, agents, reasoning, facts and long documents. Mercury 2 wins following instructions, and costs 94% less.
Which is better for coding?
Claude Sonnet 4.5. It wins 4 of the 4 coding tests both models report; Mercury 2 wins none.
Which is cheaper?
Claude Sonnet 4.5 costs $3.00 per million input tokens and $15.00 per million output tokens; Mercury 2 costs $0.25 and $0.75. That makes Mercury 2 about 94% cheaper for the same work.
How do you compare the two?
We use the 25 benchmark tests both models have published scores on. The verdict counts the 15 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 10 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.