Claude Opus 4.8 vs GPT-5.4
Full profiles: Claude Opus 4.8 (Anthropic) and GPT-5.4 (OpenAI): every benchmark, price and rank.
Updated · Data from 13 sources
As of 12 Nov 2026, Claude Opus 4.8 and GPT-5.4 are neck and neck on quality (60.4 vs 63.1 on the BenchLeader Index, inside the ±6.2 margin), and GPT-5.4 is 1.8× cheaper at $5.63 vs $10.00 per 1M tokens blended. Choose Claude Opus 4.8 for multimodal (64 vs 61) and maths (68 vs 65) and fast first answers (0.8 s vs 111.3 s); choose GPT-5.4 for reasoning (66 vs 60) and coding (65 vs 61).
Key facts (as of 12 Nov 2026)
- Claude Opus 4.8 scores 60.4 on the BenchLeader Index and GPT-5.4 scores 63.1 (BenchLeader, 12 Nov 2026).
- Claude Opus 4.8 costs $10.00 per 1M tokens blended (list price, 12 Nov 2026).
- GPT-5.4 costs $5.63 per 1M tokens blended (list price, 12 Nov 2026).
- Of the 3 benchmarks both models have, GPT-5.4 leads 2 (BenchLeader, 12 Nov 2026).
Which to pick for…
The pick's numbers come first. Category scores on the BenchLeader scale (50 is the average model, 15 points one standard deviation). High confidence means a lead of half a standard deviation or more; under one point is called either.
| Use case | Pick | Why | Confidence |
|---|---|---|---|
| Coding | GPT-5.4 | 65 vs 61 category score; DeepSWE v1.1 51.8% vs 51.8% | Medium |
| Maths | Claude Opus 4.8 | 68 vs 65 category score | Medium |
| Reasoning | GPT-5.4 | 66 vs 60 category score; ARC-AGI-2 74.0% vs 72.1% | Medium |
| Images and other media | Claude Opus 4.8 | 64 vs 61 category score | Medium |
| Chat quality (human preference) | – | Only Claude Opus 4.8 has an Arena rating (67 category score) | Not comparable |
| Lowest cost | GPT-5.4 | $5.63 vs $10.00 per 1M blended; agentic coding session $0.210 vs $0.400 | – |
| Fastest first answer | Claude Opus 4.8 | 0.8 s (Claude Opus 4.8 (high)) vs 111.3 s (GPT-5.4 (xhigh)) to the first answer token | – |
| High volume on a budget | Neither flagship | A smaller model is usually the better fit: Cheapest good LLMs, Fastest LLMs | – |
Key differences
- Claude Opus 4.8 (high) and GPT-5.4 (xhigh) are neck and neck on quality (60.4 vs 63.1, within the ±6.2 margin).
- Claude Opus 4.8 (high) is stronger in maths, multimodal.
- GPT-5.4 (xhigh) is stronger in coding, reasoning.
- GPT-5.4 (xhigh) is 1.8× cheaper ($5.63 vs $10.00 per 1M blended).
At a glance
| Spec | Claude Opus 4.8 | GPT-5.4 |
|---|---|---|
| Developer | Anthropic | OpenAI |
| Released | 28 May 2026 | 5 Mar 2026 |
| Context window (tokens) | 1,000,000 | 1,050,000 |
| Compared at | Claude Opus 4.8 (high) | GPT-5.4 (xhigh) |
| Open weights | No | No |
| Price in / out / cached input per 1M | $5.00 / $25.00 / $0.500 | $2.50 / $15.00 / $0.250 |
Category by category
Category scores on the BenchLeader scale: 50 is the average evaluated model, 15 points is one standard deviation. A gap under one point is called level.
| Category | Claude Opus 4.8 (high) | GPT-5.4 (xhigh) | Leader |
|---|---|---|---|
| Reasoning | 59.8 | 65.6 | GPT-5.4 |
| Coding | 61.0 | 64.7 | GPT-5.4 |
| Agents & tools | – | 63.0 | not comparable |
| Maths | 67.5 | 65.0 | Claude Opus 4.8 |
| Knowledge | – | 58.6 | not comparable |
| Instruction following | – | 72.4 | not comparable |
| Human preference | 67.2 | – | not comparable |
| Multimodal | 64.0 | 60.9 | Claude Opus 4.8 |
| Long context | – | 65.7 | not comparable |
On one chart
Each domain on the index's own scale, where 50 is an average model. Click a domain to see the benchmarks behind it. Pick up to 5 models from the list.
Not drawn, measured for fewer than two of these models: Agents & tools, Knowledge, Instruction following, Long context, Human preference.
Where sources disagree
Across the 3 benchmarks both have, Claude Opus 4.8 leads 0, GPT-5.4 leads 2 and 1 is a tie. Every overall index that rates both puts Claude Opus 4.8 ahead.
| Overall index | Claude Opus 4.8 (high) | GPT-5.4 (xhigh) | Leader |
|---|---|---|---|
| BenchLeader Index | 60.4 | 63.1 | neck and neck |
| Arena Text | 1482 | 1475† | Claude Opus 4.8 |
Each publisher runs its own tasks, harnesses and effort settings, so their overall verdicts can differ; BenchLeader reconciles them on one scale. † measured at another effort setting of the same model (the publisher did not test the one compared here).
Price, speed and latency
List prices blended at three input tokens per output token. Per-task prices use the token counts on the methodology page. Cost, tokens and time per benchmark task are measured: Artificial Analysis’ averages over its whole Intelligence Index run.
| Measure | Claude Opus 4.8 (high) | GPT-5.4 (xhigh) |
|---|---|---|
| Blended price per 1M tokens | $10.00 | $5.63 |
| Input / output per 1M | $5.00 / $25.00 | $2.50 / $15.00 |
| Cached input per 1M | $0.500 | $0.250 |
| Price per task: chat reply | $0.0095 | $0.0055 |
| Price per task: summarize a 30-page report | $0.075 | $0.039 |
| Price per task: code edit | $0.068 | $0.037 |
| Price per task: agentic coding session | $0.400 | $0.210 |
| Price per task: structured extraction | $0.015 | $0.0080 |
| Agentic coding session, input mostly cached | $0.198 | $0.109 |
| Output speed | 68 tok/s | 68 tok/s |
| Time to first answer | 0.8 s | 111.3 s |
| Time to a complete answer | – | 188.0 s |
Each model is compared at its best-ranked reasoning-effort setting. Claude Opus 4.8 (high) scores 60.4 and takes 1 s to its first answer token; Claude Opus 4.8 (max) scores 59.9 in 36 s, Claude Opus 4.8 scores 59.6 in 1 s. GPT-5.4 (xhigh) scores 63.1 and takes 111 s to its first answer token; GPT-5.4 (low) scores 58.8 in 3 s, GPT-5.4 (high) scores 58.7 in 2 s. Higher effort means more thinking before the answer: better scores, longer waits and more tokens billed. How reasoning effort changes scores and cost.
Every benchmark
143 benchmarks, grouped by category; each name links to the board and its leaders. † measured at another effort setting of the same model, shown for reference and not counted as a win.
| Metric | Publisher | Claude Opus 4.8 (high) | GPT-5.4 (xhigh) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 60.4 | 63.1 |
| Coding score | BenchLeader | 61.0 | 64.7 |
| Human preference score | BenchLeader | 67.2 | – |
| Maths score | BenchLeader | 67.5 | 65.0 |
| Multimodal score | BenchLeader | 64.0 | 60.9 |
| Reasoning score | BenchLeader | 59.8 | 65.6 |
| Agents & tools score | BenchLeader | – | 63.0 |
| Instruction following score | BenchLeader | – | 72.4 |
| Knowledge score | BenchLeader | – | 58.6 |
| Long context score | BenchLeader | – | 65.7 |
| Blended price $/M | BenchLeader | $10.00 | $5.63 |
| Output speed | BenchLeader | 68 tok/s | 68 tok/s |
| Time to first answer | BenchLeader | 0.8 s | 111.3 s |
| Context window | BenchLeader | 1,000,000 tokens | 1,050,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 91.0%† | 93.3% |
| Humanity's Last Exam | Scale AI / CAIS | – | 36.2% |
| SimpleBench | SimpleBench | 64.8%† | – |
| Arena Hard Prompts | Arena (LMArena) | 1513 | 1487† |
| LiveBench 1.0 Reasoning | LiveBench | 89.2%† | 88.1% |
| GPQA Diamond (AA) | Artificial Analysis | 92.0%† | 92.0% |
| Humanity's Last Exam (AA) | Artificial Analysis | 48.7%† | 43.7% |
| GPQA Diamond (Vals) | Vals AI | 92.4%† | 91.7% |
| EnigmaEval | Scale AI SEAL | 23.5%† | – |
| Kagi LLM Benchmark | Kagi LLM Benchmark | 88.8%† | 63.8%† |
| ARC-AGI-1 | ARC Prize | 92.0% | 93.7% |
| ARC-AGI-2 | ARC Prize | 72.1% | 74.0% |
| ARC-AGI-3 | ARC Prize | 1.5% | 0.2%† |
| CritPt | Artificial Analysis | 20.9%† | 23.4% |
| Chess Puzzles | Epoch AI Benchmarking Hub | 52.0%† | 48.0% |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 36.0%† | 37.0% |
| CL-bench | Epoch AI Benchmarking Hub | – | 27.9% |
| CL-bench Life | Epoch AI Benchmarking Hub | – | 21.7% |
| LMCA | Epoch AI Benchmarking Hub | 57.5%† | 52.0% |
| DTBench | Epoch AI Benchmarking Hub | 94.9%† | 94.4% |
| ForecastBench | ForecastBench | 62.4%† | 63.0%† |
| BTF-3 | Epoch AI Benchmarking Hub | 14.0% | – |
| Coding | |||
| SWE-bench Verified (Epoch) | Epoch AI Benchmarking Hub | – | 76.9%† |
| SciCode | SciCode | 53.5%† | 56.6% |
| WeirdML | WeirdML | 70.5%† | 77.7% |
Show all 143 benchmarks
| Metric | Publisher | Claude Opus 4.8 (high) | GPT-5.4 (xhigh) |
|---|---|---|---|
| Coding | |||
| FrontierCode | Cognition | 46.5%† | – |
| GSO-Bench | GSO-Bench | 47.1%† | 31.4% |
| Arena Coding | Arena (LMArena) | 1533 | 1511† |
| Arena WebDev | Arena (LMArena) | 1556 | 1399† |
| LiveBench 1.0 Coding | LiveBench | 81.8%† | 77.5% |
| SciCode (AA) | Artificial Analysis | 54.4%† | – |
| LiveCodeBench | Vals AI | 87.8%† | 84.1% |
| SWE-bench (Vals) | Vals AI | 88.6%† | 78.2% |
| SWE-Bench Pro | Scale AI SEAL | – | 59.1% |
| Code Migration | Vals AI | 47.3%† | 35.0% |
| ProgramBench | Vals AI | 1.0%† | 0.0% |
| Vibe Code Bench v1.1 | Vals AI | 82.7%† | 67.4% |
| SWE-Bench Pro (private) | Scale AI SEAL | – | 59.1% |
| SWE Atlas: Codebase QnA | Scale AI SEAL | 57.3%† | 40.8% |
| SWE Atlas: Refactoring | Scale AI SEAL | 46.7%† | 44.3% |
| SWE Atlas: Test Writing | Scale AI SEAL | 49.6%† | 44.4% |
| MirrorCode | Epoch AI Benchmarking Hub | 65.6%† | 23.3% |
| Surface Evolver Bench | Epoch AI Benchmarking Hub | 87.5% | – |
| DeepSWE v1.1 | Epoch AI Benchmarking Hub | 51.8% | 51.8% |
| ALE-Bench | Epoch AI Benchmarking Hub | 1563.8 | 1607† |
| AlgoTune | AlgoTune | – | 1.9† |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 21.7%† | – |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 84.6%† | 78.3% |
| GameBench 2 Overall | DuelLab GameBench 2 | 49.6† | 54.0† |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | 46.0† | 62.2 |
| Design Arena Website | Design Arena | 1254† | 1193† |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 23.6%† | 81.8%† |
| OSWorld-Verified 2.0 | OSWorld | 20.6%† | – |
| Remote Labor Index | Scale AI / CAIS | 8.3%† | – |
| APEX-Agents | Mercor | 48.9%† | 52.4%† |
| LiveBench Agentic Coding | LiveBench | 50.5%† | 53.8% |
| Terminal-Bench Hard | Artificial Analysis | 58.3%† | 57.6% |
| τ²-Bench Telecom (AA) | Artificial Analysis | 94.4%† | 87.1% |
| Terminal-Bench 2.1 (Vals) | Vals AI | 71.9%† | – |
| MCP Atlas | Scale AI SEAL | 82.2%† | 70.6% |
| HiL-Bench | Scale AI SEAL | 35.3%† | 9.7%† |
| τ³-bench Banking | τ²-bench | 39.7%† | 39.4% |
| τ³-Banking (AA) | Artificial Analysis | 34.2%† | 39.6% |
| Analyst Agent (AA) | Artificial Analysis | 45.0%† | – |
| APEX-Agents (AA) | Artificial Analysis | – | 33.3% |
| Finance Agent v2 | Vals AI | 53.9%† | – |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 9.6%† | 0.0% |
| Legal Research Bench | Vals AI | 43.8%† | – |
| SkillsBench | Vals AI | 59.2%† | 51.7% |
| Tax Agent Bench | Vals AI | 30.0%† | – |
| Terminal-Bench 2.0 (Vals) | Vals AI | 70.0%† | 58.4%† |
| Terminal-Bench 4.0 (Vals) | Vals AI | 23.2%† | – |
| Terminal-Bench Science | Vals AI | 4.3%† | – |
| Time Horizon Index: KSP | Vals AI | 11.8%† | – |
| Arena Search | Arena (LMArena) | 1194† | 1196† |
| DeepResearch Bench | DeepResearch Bench | 50.2% | 35.1%† |
| PostTrainBench | Epoch AI Benchmarking Hub | 30.7% | 17.8%† |
| METR Time Horizons | METR | – | 74.3% |
| GBAEval | Epoch AI Benchmarking Hub | 70.9%† | 45.1%† |
| Vending-Bench 2 | Andon Labs | 5787.4† | 6144.2† |
| AutomationBench | Artificial Analysis | 45.6%† | – |
| GDP.pdf (AA) | Artificial Analysis | 22.8%† | – |
| EnterpriseOps-Gym | Artificial Analysis | 44.0%† | – |
| AA-Briefcase v1.1 | Artificial Analysis | 1320† | – |
| Design Arena Agentic | Design Arena | 1219† | 1035† |
| DeepSearchQA | – | 63.7%† | |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 80.0%† | 78.6% |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 56.1%† | 49.0% |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 98.3%† | 95.3% |
| ProofBench | Vals AI | 69.0%† | 56.0% |
| LiveBench Mathematics | LiveBench | 94.3%† | 94.2% |
| AIME (Vals) | Vals AI | – | 96.7% |
| AIME 2026 | MathArena | 100.0%† | 99.2% |
| HMMT February 2026 | MathArena | 95.5%† | 97.7% |
| USAMO 2026 | MathArena | – | 95.2% |
| MathArena Apex | MathArena | 81.3%† | 54.2% |
| Arena Maths | Arena (LMArena) | 1492 | 1459† |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 53.0%† | 45.1% |
| LiveBench 1.0 Data Analysis | LiveBench | 66.0%† | 79.3% |
| MMLU-Pro | Vals AI | 89.6%† | 87.5% |
| LegalBench | Vals AI | 83.6%† | 86.0% |
| CorpFin | Vals AI | 66.7%† | 65.3% |
| TaxEval | Vals AI | 75.6%† | 74.0% |
| MedQA | Vals AI | – | 96.1% |
| PRBench Finance | Scale AI SEAL | – | 45.6%† |
| PRBench Legal | Scale AI SEAL | – | 44.4%† |
| MultiNRC | Scale AI SEAL | – | 58.3%† |
| CaseLaw v2 | Vals AI | – | 63.8% |
| Excel Modeling Benchmark | Vals AI | 69.4%† | – |
| MedCode | Vals AI | 53.2%† | 41.3% |
| MedScribe | Vals AI | 85.8%† | 77.5% |
| Public Benefits Bench v1.1 | Vals AI | 68.1%† | – |
| DrugDiscoveryBench | Scale AI SEAL | 46.8%† | – |
| EBR-bench | Epoch AI Benchmarking Hub | 28.6%† | 25.4% |
| GDP.pdf | Epoch AI Benchmarking Hub | 24.0%† | – |
| Instruction following | |||
| LiveBench Language | LiveBench | 79.7%† | 82.6% |
| LiveBench Instruction Following | LiveBench | 72.0%† | 70.2% |
| IFBench | Artificial Analysis | 62.2%† | 74.0% |
| Arena Instruction Following | Arena (LMArena) | 1492 | 1460† |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1482 | 1465† |
| EQ-Bench 4 | EQ-Bench | 1268† | 1258† |
| Arena Creative Writing | Arena (LMArena) | 1470 | 1433† |
| Arena Multi-turn | Arena (LMArena) | 1497 | 1480† |
| Arena Longer Queries | Arena (LMArena) | 1503 | 1477† |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1295 | 1292† |
| MMMU-Pro | Artificial Analysis | – | 78.4% |
| VISTA | Scale AI SEAL | – | 50.9%† |
| MMMU-Pro (Vals) | Vals AI | 86.6%† | 87.5% |
| MortgageTax | Vals AI | 69.9%† | 68.3% |
| VTB | Scale AI SEAL | – | 29.2%† |
| GeoBench | GeoBench | 81.0% | – |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 14.5%† | 27.1%† |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 77.7%† | 82.0% |
| Arena Document | Arena (LMArena) | 1485 | 1471† |
| MLCR | Artificial Analysis | 19.4%† | – |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 159.9† | 157.8† |
| LiveBench 1.0 | LiveBench | 76.2%† | 78.0% |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 41.8† | 39.0 |
| AA Progress Index v1 | Artificial Analysis | 318.4† | 311.6 |
| Vals RSI Index | Vals AI | 24.0%† | 15.5% |
| Vals Multimodal Index | Vals AI | 70.9%† | – |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | 18.2%† | – |
| MASK | Scale AI SEAL | – | 89.7%† |
Data as of 2026-11-12. Best configuration of each model; every score links to its source on the model pages.
DuelLab GameBench 2, release gb2-20261110-2acac640a228 (https://benchmarks.duellab.org/releases/gb2-20261110-2acac640a228/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.
What changed
- Claude Opus 4.8 (high): 1 score revision.
- Claude Opus 4.8 (high): 1 new result (GeoBench).
- GPT-5.4 (xhigh): 1 new result (MirrorCode).
- GPT-5.4 (xhigh): BenchLeader Index 63.7 → 63.2.
- GPT-5.4 (xhigh): 1 score revision.
- GPT-5.4 (xhigh): 1 new result (AA Progress Index v1).
- GPT-5.4 (xhigh): 2 new results (Harvey's Legal Agent Benchmark v1, Vals RSI Index).
- Claude Opus 4.8 (high): BenchLeader Index 61.1 → 60.6.
- GPT-5.4 (xhigh): 2 new results (USAMO 2026, τ³-bench Banking).
- Claude Opus 4.8 (high): 1 score revision.
Related comparisons
- Claude Opus 4.8 vs Gemini 4 Carbon
- GPT-5.4 vs Claude Fable 5.2
- GPT-5.4 vs Gemini 4 Carbon
- GPT-5.4 vs Claude Opus 5.5
- GPT-5.4 vs GPT-6 Astra
- Claude Opus 4.7 vs GPT-5.4
- Claude Opus 4.8 vs GPT-6.1 Sol
- Claude Opus 4.8 vs GPT-5.6 Sol
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Claude Opus 4.8 vs GPT-5.4: questions
- Is Claude Opus 4.8 better than GPT-5.4?
- Claude Opus 4.8 (high) and GPT-5.4 (xhigh) are neck and neck on quality (60.4 vs 63.1, within the ±6.2 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 12 Nov 2026, so check the category scores for your use.
- Claude Opus 4.8 vs GPT-5.4: which is cheaper, and will that change?
- GPT-5.4 is cheaper today: $5.63 against $10.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is Claude Opus 4.8 or GPT-5.4 better for coding?
- GPT-5.4 scores higher in coding (65 vs 61 on the category index, where 50 is average).
- Which is faster, Claude Opus 4.8 or GPT-5.4?
- GPT-5.4 streams faster (68 vs 68 output tokens per second, Artificial Analysis 7-day median).
- Which is better for long documents?
- Both take about 1.1M tokens: Claude Opus 4.8 1,000,000, GPT-5.4 1,050,000.
More questions (3)
- Which has lower latency?
- Claude Opus 4.8 answers first: 0.8 s against 111.3 s to the first answer token, as Artificial Analysis measures it. Reasoning effort drives most of this: a higher setting thinks longer before it answers.
- Do independent benchmarks agree on Claude Opus 4.8 vs GPT-5.4?
- Across the 3 benchmarks both have, Claude Opus 4.8 leads 0, GPT-5.4 leads 2 and 1 is a tie. Every overall index that rates both puts Claude Opus 4.8 ahead.
- When was Claude Opus 4.8 released?
- Claude Opus 4.8 was released on 28 May 2026.