Agnes 2.5 Pro Beta vs Claude Sonnet 4
Wins 5 of 6 areas
Coding · Agents · Reasoning · Images and charts · Long documents
Wins 1 of 6 areas
Facts
Agnes 2.5 Pro Beta is the stronger all-rounder.Claude Sonnet 4 is better at facts.
Scores updated · 15 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.
- ReasoningHard problems that need careful thinking30Agnes 2.5 Pro Beta3 of 3 tests
- CodingWriting and fixing software20Agnes 2.5 Pro Beta2 of 2 tests
- AgentsCarrying out multi-step tasks on its own20Agnes 2.5 Pro Beta2 of 2 tests
- Images and chartsUnderstanding pictures, charts and video10Agnes 2.5 Pro Beta1 of 1 test
- Long documentsFinding answers in very long texts10Agnes 2.5 Pro Beta1 of 1 test
- FactsGetting facts right instead of making them up02Claude Sonnet 42 of 2 tests
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.
Agnes 2.5 Pro Beta costs 98% 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 Agnes 2.5 Pro Beta pulls ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+33.4points ahead
- Real work tasks from 44 professionsGDPVal+31.8points ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+26.8points ahead
Where Claude Sonnet 4 pulls ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+5.9points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+3.7points ahead
Every test, side by side
All 15 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingAgnes 2.5 Pro Beta
- Terminal-Bench 2.1Agnes 2.5 Pro Beta by 33.469.736.3+33.4
- SciCodeAgnes 2.5 Pro Beta by 8.848.840+8.8
AgentsAgnes 2.5 Pro Beta
- GDPValAgnes 2.5 Pro Beta by 31.841.49.6+31.8
- τ-Bench V3 · BankingAgnes 2.5 Pro Beta by 1935.716.7+19
ReasoningAgnes 2.5 Pro Beta
- Humanity's Last ExamAgnes 2.5 Pro Beta by 26.837.510.7+26.8
- CritPtAgnes 2.5 Pro Beta by 15.415.70.3+15.4
- GPQA DiamondAgnes 2.5 Pro Beta by 12.890.577.7+12.8
FactsClaude Sonnet 4
- AA-Omniscience · AccuracyClaude Sonnet 4 by 5.916.822.7+5.9
- AA-Omniscience · Non-hallucinationClaude Sonnet 4 by 3.767.270.9+3.7
Images and chartsAgnes 2.5 Pro Beta
- MMMU-ProAgnes 2.5 Pro Beta by 13.675.461.8+13.6
Long documentsAgnes 2.5 Pro Beta
- AA-LCRAgnes 2.5 Pro Beta by 12.78370.3+12.7
Other results4 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 Agentic IndexAgnes 2.5 Pro Beta by 26.243.817.6+26.2
- Artificial Analysis Coding IndexAgnes 2.5 Pro Beta by 24.762.337.6+24.7
- AA IntelligenceAgnes 2.5 Pro Beta by 16.335.218.9+16.3
- AA-OmniscienceClaude Sonnet 4 by 10.7-10.50.2+10.7
Questions people ask
Which is better, Agnes 2.5 Pro Beta or Claude Sonnet 4?
Agnes 2.5 Pro Beta wins five of the six areas where both have results: coding, agents, reasoning, images and charts and long documents. Claude Sonnet 4 wins facts.
Which is better for coding?
Agnes 2.5 Pro Beta. It wins 2 of the 2 coding tests both models report; Claude Sonnet 4 wins none.
Which is cheaper?
Agnes 2.5 Pro Beta costs $0.10 per million input tokens and $0.30 per million output tokens; Claude Sonnet 4 costs $3.00 and $15.00. That makes Agnes 2.5 Pro Beta about 98% cheaper for the same work.
How do you compare the two?
We use the 15 benchmark tests both models have published scores on. The verdict counts the 11 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 4 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.