Claude vs ChatGPT: AI benchmark scores compared (Nov 2026)
On the 21 public leaderboards where both have a reported score, Claude’s best-ranked model scores higher on 16, ChatGPT’s on 5. Counting only the 12 leaderboards their sources still update: Claude 10, ChatGPT 2.
These counts compare Claude’s and ChatGPT’s best-ranked model on each leaderboard, so the model version can vary by benchmark (most often Claude Opus 5 for Claude and GPT-6 Astra for ChatGPT). See the table below for the exact model on each.
Free to use on your site, blog or newsletter. Please keep the “via chatjudge.ai” credit and the source link under the chart. The chart updates itself when we refresh the scores.
Citation
Claude vs ChatGPT on 21 public AI leaderboards: Claude’s best-ranked model scored higher on 16, ChatGPT’s on 5. The model version varies by benchmark. Original sources: Vals AI, Legal Benchmarks, Artificial Analysis, Datacurve, Scale AI and Center for AI Safety, LMArena (leaderboard dates per row on the page). Via ChatJudge: https://chatjudge.ai/compare/claude-vs-chatgpt, checked 17 Nov 2026.
In plain words
- Multi-domain agent tasks: Claude higher on 1 of 1 (Vals Index).
- Legal: Claude higher on all 4 (Legal Benchmarks: Contract Workflows, Legal Benchmarks: Data Extraction, LegalBench and Legal Research Bench).
- Health & medical: Claude higher on 2 of 3 (HealthBench (independent run) and MedCode); ChatGPT higher on 1 of 3 (MedQA).
- Finance: Claude higher on all 2 (Finance Agent and CorpFin v2).
- Coding: Claude higher on 2 of 3 (SWE-bench Verified and LiveCodeBench); ChatGPT higher on 1 of 3 (DeepSWE).
- Agents & tool use: Claude higher on all 2 (Terminal-Bench 4.0 and Terminal-Bench 2.1).
- Reasoning & knowledge: Claude higher on 1 of 4 (MMLU Pro); ChatGPT higher on 3 of 4 (Humanity's Last Exam, GPQA Diamond and AIME).
- Chat preference: Claude higher on 1 of 1 (LMArena Text Arena).
- Multilingual: Claude higher on 1 of 1 (MGSM).
- Claude’s best-ranked entry is most often Claude Opus 5 (5 of 21); ChatGPT’s is most often GPT-6 Astra (5 of 21).
Claude here means Anthropic’s Claude models; ChatGPT means OpenAI’s GPT and o-series models (the open-weight GPT OSS models are left out). On each leaderboard we compare each side’s best-ranked model, so the model can differ by benchmark. A side only counts where it appears on the leaderboard (we keep each source’s top rows).
Results from published AI leaderboards. Compare models within each benchmark; scores from different tests aren’t directly comparable. How we collect these results
Multi-domain agent tasks
| Benchmark | Claude | ChatGPT | Higher score |
|---|---|---|---|
| Vals IndexAccuracy · dated 15 Nov 2026 | 71.31%Claude Fable 5.2 · #1 | 65.02%GPT-6 Astra · #8 | Claude higherby 6.29 pts |
Legal
| Benchmark | Claude | ChatGPT | Higher score |
|---|---|---|---|
| Legal Benchmarks: Contract WorkflowsTask pass rate · dated Nov 2026 | 48.4%Claude Opus 5.5 · #1 | 25.8%GPT-6 Astra · #10 | Claude higherby 22.6 pts |
| Legal Benchmarks: Data ExtractionTask pass rate · dated Nov 2026 | 76.7%Claude Opus 5.5 · #1 | 56.7%GPT-6 Astra · #4 | Claude higherby 20.0 pts |
| LegalBenchAccuracy · dated 13 Oct 2026 | 88.56%Claude Fable 5 · #1 | 86.97%GPT-5.6 Sol · #10 | Claude higherby 1.59 pts |
| Legal Research BenchAll-Pass · dated 11 Nov 2026 | 55.29%Claude Opus 5 · #3 | 48.08%GPT-5.6 Sol · #14 | Claude higherby 7.21 pts |
Health & medical
| Benchmark | Claude | ChatGPT | Higher score |
|---|---|---|---|
| HealthBench (independent run)HealthBench · checked 17 Nov 2026 | 62.1%Claude Sonnet 5.5 (max) · #1 | 58.5%GPT-6.1 Sol (max) · #4 | Claude higherby 3.6 pts |
| MedCodeAccuracy · dated 29 Oct 2026 | 63.57%Claude Opus 5 · #1 | 52.73%GPT 5.1 · #13 | Claude higherby 10.84 pts |
| MedQAAccuracy · dated 16 Apr 2026 · no longer updated | 95.88%Claude Opus 4.5 (Thinking) · #10 | 96.52%o1 · #1 | ChatGPT higherby 0.64 pts |
Finance
| Benchmark | Claude | ChatGPT | Higher score |
|---|---|---|---|
| Finance AgentAccuracy · dated 11 Nov 2026 | 69.47%Claude Fable 5.2 · #1 | 61.82%GPT-6.1 Sol · #20 | Claude higherby 7.65 pts |
| CorpFin v2Accuracy · dated 12 Aug 2026 · no longer updated | 73.19%Claude Opus 5 · #1 | 68.42%GPT 5.5 · #9 | Claude higherby 4.77 pts |
Coding
| Benchmark | Claude | ChatGPT | Higher score |
|---|---|---|---|
| DeepSWEPass@1 · dated 22 Sep 2026 | 73.65% ±3.87Claude Opus 5 (max) · #3 | 74.12% ±2.87GPT-6 Astra (xhigh) · #1 | ChatGPT higherby 0.47 pts, within ± |
| SWE-bench VerifiedScore · dated 1 Sep 2026 · no longer updated | 97.0%Claude Opus 5 · #1 | 96.2%GPT-5.6 Sol · #3 | Claude higherby 0.8 pts |
| LiveCodeBenchAccuracy · dated 1 Sep 2026 · no longer updated | 90.52%Claude Fable 5.1 · #1 | 87.99%GPT 5.2 Codex · #9 | Claude higherby 2.53 pts |
Agents & tool use
| Benchmark | Claude | ChatGPT | Higher score |
|---|---|---|---|
| Terminal-Bench 4.0Score · dated 11 Nov 2026 | 65.15%Claude Fable 5.2 · #1 | 59.60%GPT-6 Astra · #6 | Claude higherby 5.55 pts |
| Terminal-Bench 2.1Score · dated 1 Oct 2026 · no longer updated | 87.64%Claude Opus 5.5 · #1 | 87.27%GPT-6 Astra · #2 | Claude higherby 0.37 pts |
Reasoning & knowledge
| Benchmark | Claude | ChatGPT | Higher score |
|---|---|---|---|
| Humanity's Last ExamAccuracy · checked 17 Nov 2026 | 46.50% ±2Fable 5.1 (xhigh) · #2 | 54.80% ±1.94GPT 6 Astra · #1 | ChatGPT higherby 8.30 pts |
| GPQA DiamondAccuracy · dated 1 Sep 2026 · no longer updated | 93.43%Claude Opus 5 · #8 | 95.20%GPT-5.6 Sol · #2 | ChatGPT higherby 1.77 pts |
| AIMEAccuracy · dated 16 Apr 2026 · no longer updated | 96.25%Claude Opus 4.7 · #7 | 96.88%GPT 5.2 · #2 | ChatGPT higherby 0.63 pts |
| MMLU ProAccuracy · dated 1 Sep 2026 · no longer updated | 92.38%Claude Fable 5.1 · #1 | 89.10%GPT-5.6 Sol · #16 | Claude higherby 3.28 pts |
Chat preference
| Benchmark | Claude | ChatGPT | Higher score |
|---|---|---|---|
| LMArena Text ArenaArena score · dated 11 Nov 2026 | 1507 ±6Claude Opus 5.5 (High) · #4 | 1485 ±4GPT-5.6 Sol (xHigh) · #23 | Claude higherby 22 |
Multilingual
Every score links to the leaderboard that published it. Sources: Vals AI ↗ · Legal Benchmarks ↗ · Artificial Analysis ↗ · Datacurve ↗ · Scale AI and Center for AI Safety ↗ · LMArena ↗. ChatJudge does not run these tests.
Test Claude and ChatGPT on your own work
Run the same prompt in both and compare the answers. Prompts for Claude → · Prompts for ChatGPT →
Questions people ask
Is Claude better than ChatGPT?
On the 21 public leaderboards where both have a reported score, Claude’s best-ranked model scores higher on 16, ChatGPT’s on 5. Counting only the 12 leaderboards their sources still update: Claude 10, ChatGPT 2. Which is better depends on the task: the leaders differ by benchmark, and scores from different tests are not directly comparable.
Is Claude or ChatGPT better for multi-domain agent tasks?
Vals Index: Claude 71.31% vs ChatGPT 65.02%. Each figure is the best-ranked model from each side on that leaderboard, as published by its source.
Is Claude or ChatGPT better for legal work?
Legal Benchmarks: Contract Workflows: Claude 48.4% vs ChatGPT 25.8%; Legal Benchmarks: Data Extraction: Claude 76.7% vs ChatGPT 56.7%; LegalBench: Claude 88.56% vs ChatGPT 86.97%; Legal Research Bench: Claude 55.29% vs ChatGPT 48.08%. Each figure is the best-ranked model from each side on that leaderboard, as published by its source.
Is Claude or ChatGPT better for health & medical questions?
HealthBench (independent run): Claude 62.1% vs ChatGPT 58.5%; MedCode: Claude 63.57% vs ChatGPT 52.73%; MedQA: ChatGPT 96.52% vs Claude 95.88%. Each figure is the best-ranked model from each side on that leaderboard, as published by its source.
Is Claude or ChatGPT better for finance?
Finance Agent: Claude 69.47% vs ChatGPT 61.82%; CorpFin v2: Claude 73.19% vs ChatGPT 68.42%. Each figure is the best-ranked model from each side on that leaderboard, as published by its source.
Is Claude or ChatGPT better for coding?
DeepSWE: ChatGPT 74.12% vs Claude 73.65% (within the published ± margins); SWE-bench Verified: Claude 97.0% vs ChatGPT 96.2%; LiveCodeBench: Claude 90.52% vs ChatGPT 87.99%. Each figure is the best-ranked model from each side on that leaderboard, as published by its source.
How recent is this comparison?
It is rebuilt from our benchmark data, last checked on 17 Nov 2026. Each leaderboard’s own date is shown in the table. MedQA, CorpFin v2, SWE-bench Verified, LiveCodeBench, Terminal-Bench 2.1, GPQA Diamond, AIME, MMLU Pro and MGSM are no longer updated by the source, so newer models may be missing there.
Which models are being compared?
On each leaderboard we take the best-ranked Claude model and the best-ranked ChatGPT model, so the model can change from one benchmark to the next. Claude’s top entry is most often Claude Opus 5 (5 of 21); ChatGPT’s is most often GPT-6 Astra (5 of 21). Claude here means Anthropic’s Claude models; ChatGPT means OpenAI’s GPT and o-series models (the open-weight GPT OSS models are left out).
