Claude Sonnet 4.5 vs GPT-5.1
Wins 1 of 7 areas
Agents
Wins 5 of 7 areas
Reasoning · Facts · Images and charts · Long documents · Following instructions
GPT-5.1 is the stronger all-rounder.Claude Sonnet 4.5 is better at agents.
Scores updated · 51 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 thinking14GPT-5.14 of 5 tests
- Images and chartsUnderstanding pictures, charts and video02GPT-5.12 of 2 tests
- Following instructionsDoing exactly what it is asked02GPT-5.12 of 2 tests
- FactsGetting facts right instead of making them up12GPT-5.12 of 4 tests · 1 tie
- Long documentsFinding answers in very long texts01GPT-5.11 of 1 test
- AgentsCarrying out multi-step tasks on its own21Claude Sonnet 4.52 of 3 tests
- CodingWriting and fixing software22Even2 each · 2 ties
Long documents rests on a single test.
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.
GPT-5.1 costs 38% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where GPT-5.1 pulls ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+26.7points ahead
- Recent programming contest problemsLiveCodeBench v6+23points ahead
- Short factual questions, answered correctlySimpleQA Verified+17.3points ahead
Where Claude Sonnet 4.5 pulls ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+8.6points ahead
- Real work tasks from 44 professionsGDPVal+4points ahead
- Command-line tasks in a real terminalTerminal-Bench 2.1+3.4points ahead
Every test, side by side
All 51 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingEven
- LiveCodeBench v6GPT-5.1 by 236487+23
- Terminal-Bench HardGPT-5.1 by 9.935.645.5+9.9
- Terminal-Bench 2.1Claude Sonnet 4.5 by 3.455.852.4+3.4
- SciCodeClaude Sonnet 4.5 by 2.445.743.3+2.4
- SWE-bench Verifiedtie77.276.3tie
- LMArena · WebDevtie13941395tie
AgentsClaude Sonnet 4.5
- BrowseCompGPT-5.1 by 26.724.150.8+26.7
- τ-Bench V3 · BankingClaude Sonnet 4.5 by 8.624.515.9+8.6
- GDPValClaude Sonnet 4.5 by 420.516.5+4
ReasoningGPT-5.1
- Humanity's Last ExamGPT-5.1 by 10.717.828.5+10.7
- ARC-AGI-2GPT-5.1 by 413.617.6+4
- GPQA DiamondGPT-5.1 by 3.983.487.3+3.9
- CritPtGPT-5.1 by 3.81.14.9+3.8
- SimpleBenchClaude Sonnet 4.5 by 1.154.353.2+1.1
FactsGPT-5.1
- SimpleQA VerifiedGPT-5.1 by 17.330.748+17.3
- AA-Omniscience · AccuracyGPT-5.1 by 4.832.937.7+4.8
- AA-Omniscience · Non-hallucinationClaude Sonnet 4.5 by 2.850.948.1+2.8
- Vectara HHEM hallucination ratelower is bettertie1212.1tie
Images and chartsGPT-5.1
Following instructionsGPT-5.1
- IFBenchGPT-5.1 by 15.657.372.9+15.6
- Multi-ChallengeGPT-5.1 by 8.155.363.4+8.1
Other results28 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.
- HMMT Feb. 2025GPT-5.1 by 28.867.596.3+28.8
- τ²-BenchClaude Sonnet 4.5 by 15.39882.7+15.3
- vectara_avg_summary_lengthGPT-5.1 by 126.6 rating points127.8254.4+126.6 rating
- HLE (with tools)GPT-5.1 by 10.73242.7+10.7
- bbqClaude Sonnet 4.5 by 10.298.988.7+10.2
- HMMT Nov. 2025GPT-5.1 by 1081.791.7+10
- ARC-AGI-1GPT-5.1 by 9.163.772.8+9.1
- frontiermath_tier_4_v1GPT-5.1 by 8.34.212.5+8.3
- τ²-Bench (Retail)Claude Sonnet 4.5 by 8.386.277.9+8.3
- AIME 2025GPT-5.1 by 78794+7
- HarmBenchGPT-5.1 by 5.69297.6+5.6
- AA-OmniscienceGPT-5.1 by 5.5-0.15.4+5.5
- swe_bench_bashClaude Sonnet 4.5 by 5.471.466+5.4
- vectara_answer_rateGPT-5.1 by 4.495.6100+4.4
- AA Agentic IndexGPT-5.1 by 4.117.521.6+4.1
- AA IntelligenceGPT-5.1 by 420.724.7+4
- ARC-AGI-2 (Verified)GPT-5.1 by 413.617.6+4
- τ²-Bench Telecom (AA run)GPT-5.1 by 3.878.181.9+3.8
- AIR-Bench 2024Claude Sonnet 4.5 by 3.689.886.2+3.6
- Artificial Analysis Coding IndexClaude Sonnet 4.5 by 2.752.149.4+2.7
- Terminal-Bench 2.0Claude Sonnet 4.5 by 2.45047.6+2.4
- MMMLUGPT-5.1 by 1.989.191+1.9
- XSTestGPT-5.1 by 1.896.498.2+1.8
- anthropic_red_teamGPT-5.1 by 1.398.299.5+1.3
- matharena_visual_math_overallGPT-5.1 by 1.175.876.9+1.1
- MMLU-Protie87.587tie
- simple_safety_teststie10099.8tie
- vectara_factual_consistencytie8887.9tie
Questions people ask
Which is better, Claude Sonnet 4.5 or GPT-5.1?
GPT-5.1 wins five of the seven areas where both have results: reasoning, facts, images and charts, long documents and following instructions. Claude Sonnet 4.5 wins agents. They are level on coding.
Which is better for coding?
Neither. They win 2 coding tests each of the 6 both models report, and 2 are ties.
Which is cheaper?
Claude Sonnet 4.5 costs $3.00 per million input tokens and $15.00 per million output tokens; GPT-5.1 costs $1.25 and $10.00. That makes GPT-5.1 about 38% cheaper for the same work.
How do you compare the two?
We use the 51 benchmark tests both models have published scores on. The verdict counts the 23 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 28 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.