Claude Opus 4.8 vs GPT-5.6 Terra
Full profiles: Claude Opus 4.8 (Anthropic) and GPT-5.6 Terra (OpenAI): every benchmark, price and rank.
Updated · Data from 11 sources
As of 25 Oct 2026, Claude Opus 4.8 (max) and GPT-5.6 Terra (max) are neck and neck on quality (63.3 vs 64.4 on the BenchLeader Index, inside the ±6.9 margin of error), and GPT-5.6 Terra (max) costs 2.2× less, which makes it the better buy for most workloads. Pick Claude Opus 4.8 for knowledge (67 vs 57); pick GPT-5.6 Terra for instruction following (70 vs 62).
Claude Opus 4.8 benchmarks max Anthropic
GPT-5.6 Terra benchmarks max OpenAIOpen in the comparison toolWhich 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-5.6 Terra | 65 vs 59 category score; SciCode 55.0% vs 53.5% | Medium |
| Agents & tool use | GPT-5.6 Terra | 59 vs 57 category score; τ³-Banking (AA) 40.2% vs 34.2% | Medium |
| Maths | Either | 68 vs 69 category score; FrontierMath Tiers 1–3 80.0% vs 86.0% | Either |
| Reasoning | GPT-5.6 Terra | 70 vs 69 category score; GPQA Diamond 93.3% vs 91.0% | Medium |
| Knowledge | Claude Opus 4.8 | 67 vs 57 category score; SimpleQA Verified 53.0% vs 43.2% | High |
| Long documents | GPT-5.6 Terra | 67 vs 64 category score; AA-LCR 83.0% vs 77.7% | Medium |
| Following instructions | GPT-5.6 Terra | 70 vs 62 category score; IFBench 71.2% vs 62.2% | High |
| Lowest cost | GPT-5.6 Terra | $4.50 vs $10.00 per 1M blended; agentic coding session $0.168 vs $0.400 | – |
| Fastest first answer | Claude Opus 4.8 | 29.3 s (Claude Opus 4.8 (max)) vs 101.0 s (GPT-5.6 Terra (max)) 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 (max) and GPT-5.6 Terra (max) are neck and neck on quality (63.3 vs 64.4, within the ±6.9 margin).
- Claude Opus 4.8 (max) is stronger in knowledge.
- GPT-5.6 Terra (max) is stronger in agents & tools, coding, composite, instruction following, long context, maths, reasoning, multimodal.
- GPT-5.6 Terra (max) is 2.2× cheaper ($4.50 vs $10.00 per 1M blended).
- GPT-5.6 Terra (max) streams 2.0× faster (126 vs 63 tokens per second).
At a glance
| Spec | Claude Opus 4.8 | GPT-5.6 Terra |
|---|---|---|
| Developer | Anthropic | OpenAI |
| Released | 28 May 2026 | 9 Jul 2026 |
| Context window (tokens) | 1,000,000 | 1,050,000 |
| Compared at | Claude Opus 4.8 (max) | GPT-5.6 Terra (max) |
| Open weights | No | No |
| Price in / out / cached input per 1M | $5.00 / $25.00 / $0.500 | $2.00 / $12.00 / $0.200 |
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 (max) | GPT-5.6 Terra (max) | Leader |
|---|---|---|---|
| Reasoning | 68.8 | 70.1 | GPT-5.6 Terra |
| Coding | 58.9 | 65.1 | GPT-5.6 Terra |
| Agents & tools | 56.5 | 58.9 | GPT-5.6 Terra |
| Maths | 68.1 | 68.9 | level |
| Knowledge | 66.9 | 57.2 | Claude Opus 4.8 |
| Instruction following | 62.1 | 69.9 | GPT-5.6 Terra |
| Multimodal | – | 63.7 | not comparable |
| Long context | 63.8 | 66.6 | GPT-5.6 Terra |
| Composite | 76.2 | 76.6 | level |
Where sources disagree
Across the 40 benchmarks both have, Claude Opus 4.8 leads 15, GPT-5.6 Terra leads 25. Independent sources split: Claude Opus 4.8 leads on Arena Text; GPT-5.6 Terra leads on AA Intelligence Index v4.3.2.
| Overall index | Claude Opus 4.8 (max) | GPT-5.6 Terra (max) | Leader |
|---|---|---|---|
| BenchLeader Index | 63.3 | 64.4 | neck and neck |
| AA Intelligence Index v4.3.2 | 41.8 | 42.1 | GPT-5.6 Terra |
| Arena Text | 1482† | 1467 | 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; suite cost is what Artificial Analysis paid to run its whole Intelligence Index.
| Measure | Claude Opus 4.8 (max) | GPT-5.6 Terra (max) |
|---|---|---|
| Blended price per 1M tokens | $10.00 | $4.50 |
| Input / output per 1M | $5.00 / $25.00 | $2.00 / $12.00 |
| Cached input per 1M | $0.500 | $0.200 |
| Price per task: chat reply | $0.0095 | $0.0044 |
| Price per task: summarize a 30-page report | $0.075 | $0.031 |
| Price per task: code edit | $0.068 | $0.030 |
| Price per task: agentic coding session | $0.400 | $0.168 |
| Price per task: structured extraction | $0.015 | $0.0064 |
| Cost to run a full benchmark suite | $4.08 | $1.40 |
| Output speed | 63 tok/s | 126 tok/s |
| Time to first answer | 29.3 s | 101.0 s |
| Time to a complete answer | 37.2 s | 105.0 s |
Per token is not per task: on Artificial Analysis' full suite Claude Opus 4.8 (max) cost $4.08 and GPT-5.6 Terra (max) $1.40, because Claude Opus 4.8 (max) used about 1.3× as many tokens for the same work, which a per-token price hides.
Each model is compared at its best-ranked reasoning-effort setting. Claude Opus 4.8 (max) scores 63.3 and takes 29 s to its first answer token; Claude Opus 4.8 (high) scores 60.7 in 1 s, Claude Opus 4.8 scores 60.1 in 1 s. GPT-5.6 Terra (max) scores 64.4 and takes 101 s to its first answer token; GPT-5.6 Terra (xhigh) scores 62.8 in 7 s, GPT-5.6 Terra (high) scores 61.3 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
141 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 (max) | GPT-5.6 Terra (max) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 63.3 | 64.4 |
| Agents & tools score | BenchLeader | 56.5 | 58.9 |
| Coding score | BenchLeader | 58.9 | 65.1 |
| Composite score | BenchLeader | 76.2 | 76.6 |
| Instruction following score | BenchLeader | 62.1 | 69.9 |
| Knowledge score | BenchLeader | 66.9 | 57.2 |
| Long context score | BenchLeader | 63.8 | 66.6 |
| Maths score | BenchLeader | 68.1 | 68.9 |
| Reasoning score | BenchLeader | 68.8 | 70.1 |
| Multimodal score | BenchLeader | – | 63.7 |
| Blended price $/M | BenchLeader | $10.00 | $4.50 |
| Output speed | BenchLeader | 63 tok/s | 126 tok/s |
| Time to first answer | BenchLeader | 29.3 s | 101.0 s |
| Context window | BenchLeader | 1,000,000 tokens | 1,050,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 91.0% | 93.3% |
| SimpleBench | SimpleBench | 64.8%† | 48.9%† |
| Arena Hard Prompts | Arena (LMArena) | 1505† | 1490† |
| LiveBench 2.0 Reasoning | LiveBench | 89.2%† | 90.6%† |
| LiveBench 1.0 Reasoning | LiveBench | 89.2% | 90.6% |
| GPQA Diamond (AA) | Artificial Analysis | 92.0% | 92.5% |
| Humanity's Last Exam (AA) | Artificial Analysis | 48.7% | 42.9% |
| GPQA Diamond (Vals) | Vals AI | 92.4%† | 90.9%† |
| EnigmaEval | Scale AI SEAL | 23.5%† | – |
| Kagi LLM Benchmark | Kagi LLM Benchmark | 88.8%† | 51.3%† |
| ARC-AGI-1 | ARC Prize | 92.5% | 96.5% |
| ARC-AGI-2 | ARC Prize | 72.1%† | 83.9% |
| ARC-AGI-3 | ARC Prize | 1.5%† | 0.8% |
| CritPt | Artificial Analysis | 20.9% | 30.0% |
| Chess Puzzles | Epoch AI Benchmarking Hub | 52.0% | 47.0% |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 36.0% | 35.0% |
| LMCA | Epoch AI Benchmarking Hub | 57.5% | 55.0% |
| DTBench | Epoch AI Benchmarking Hub | 94.9% | 93.3% |
| ForecastBench | ForecastBench | 66.3% | – |
| BTF-3 | Epoch AI Benchmarking Hub | 14.0%† | – |
| Coding | |||
| SciCode | SciCode | 53.5% | 55.0% |
| WeirdML | WeirdML | 70.5%† | 78.3%† |
| FrontierCode | Cognition | 46.5%† | 41.3%† |
| GSO-Bench | GSO-Bench | 47.1%† | – |
| Arena Coding | Arena (LMArena) | 1529† | 1518† |
Show all 141 benchmarks
| Metric | Publisher | Claude Opus 4.8 (max) | GPT-5.6 Terra (max) |
|---|---|---|---|
| Coding | |||
| Arena WebDev | Arena (LMArena) | 1536† | 1521† |
| LiveBench 2.0 Coding | LiveBench | 77.3%† | 86.2%† |
| LiveBench 1.0 Coding | LiveBench | 81.8% | 78.3% |
| SciCode (AA) | Artificial Analysis | 54.4% | 55.0% |
| LiveCodeBench | Vals AI | 87.8%† | 85.9%† |
| IOI | Vals AI | – | 87.6% |
| SWE-bench (Vals) | Vals AI | 88.6%† | 95.4% |
| Code Migration | Vals AI | 47.3%† | 47.8% |
| ProgramBench | Vals AI | 1.0%† | 0.5% |
| Vibe Code Bench 1-100 | Vals AI | – | 14.8% |
| Vibe Code Bench v1.1 | Vals AI | 82.7%† | 74.6% |
| SWE Atlas: Codebase QnA | Scale AI SEAL | 57.3%† | – |
| SWE Atlas: Refactoring | Scale AI SEAL | 46.7%† | – |
| SWE Atlas: Test Writing | Scale AI SEAL | 49.6%† | – |
| Surface Evolver Bench | Epoch AI Benchmarking Hub | 87.5%† | 83.8%† |
| DeepSWE v1.1 | Epoch AI Benchmarking Hub | 59.0% | 69.6% |
| CursorBench | Cursor | – | 41.3% |
| ALE-Bench | Epoch AI Benchmarking Hub | 1563.8† | 1951.4 |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 21.7% | 35.4% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 84.6% | 88.0% |
| Startrise Frontend | Startrise | – | 66.9%† |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 23.6% | 21.5% |
| OSWorld-Verified 2.0 | OSWorld | 20.6% | – |
| Remote Labor Index | Scale AI / CAIS | 8.3%† | – |
| APEX-Agents | Mercor | 48.9% | 58.2% |
| Arena Agent | Arena (LMArena) | – | 0.2† |
| LiveBench Long Agentic | LiveBench | 48.3%† | 48.4%† |
| LiveBench Browser | LiveBench | 96.0%† | 84.0%† |
| LiveBench Agentic Coding | LiveBench | 50.5% | 55.0% |
| Terminal-Bench Hard | Artificial Analysis | 58.3% | 57.6% |
| τ²-Bench Telecom (AA) | Artificial Analysis | 94.4% | 86.3% |
| Terminal-Bench 2.1 (Vals) | Vals AI | 71.9%† | 77.5% |
| MCP Atlas | Scale AI SEAL | 82.2% | – |
| HiL-Bench | Scale AI SEAL | 35.3%† | – |
| τ³-bench Banking | τ²-bench | 39.7% | – |
| GDPval-AA v2.1 | Artificial Analysis | – | 47.6% |
| τ³-Banking (AA) | Artificial Analysis | 34.2% | 40.2% |
| ITBench SRE (AA) | Artificial Analysis | – | 51.0% |
| Analyst Agent (AA) | Artificial Analysis | 45.0% | – |
| APEX-Agents (AA) | Artificial Analysis | – | 38.9% |
| CyberBench | Vals AI | – | 72.1% |
| Finance Agent v2 | Vals AI | 53.9%† | 54.4% |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 9.6%† | 0.8% |
| Legal Research Bench | Vals AI | 43.8%† | 41.4% |
| SkillsBench | Vals AI | 59.2%† | 58.9% |
| SREBench | Vals AI | – | 9.5% |
| Tax Agent Bench | Vals AI | 30.0%† | 30.0% |
| Terminal-Bench 2.0 (Vals) | Vals AI | 70.0%† | – |
| Terminal-Bench 4.0 (Vals) | Vals AI | 23.2%† | 22.7% |
| Terminal-Bench Science | Vals AI | 4.3%† | 10.0% |
| Time Horizon Index: KSP | Vals AI | 11.8%† | – |
| Arena Search | Arena (LMArena) | 1195† | – |
| BALROG | BALROG | – | 53.2% |
| DeepResearch Bench | DeepResearch Bench | 50.2%† | – |
| PostTrainBench | Epoch AI Benchmarking Hub | 32.9% | – |
| GBAEval | Epoch AI Benchmarking Hub | 70.9%† | – |
| Vending-Bench 2 | Andon Labs | 2992.3 | 7343.2† |
| AutomationBench | Artificial Analysis | 45.6% | 59.6% |
| GDP.pdf (AA) | Artificial Analysis | 22.8% | 24.0% |
| EnterpriseOps-Gym | Artificial Analysis | 44.0% | 38.5% |
| AA-Briefcase v1.1 | Artificial Analysis | 1319 | – |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | – | 13.8% |
| FACTS Search | 79.6%† | 81.7%† | |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 80.0% | 86.0% |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 56.1% | 70.7% |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 98.3% | 99.7% |
| ProofBench | Vals AI | 69.0% | 74.0%† |
| LiveBench Mathematics | LiveBench | 94.3% | 94.9% |
| AIME 2026 | MathArena | 100.0% | – |
| HMMT February 2026 | MathArena | 95.5% | – |
| MathArena Apex | MathArena | 81.3% | – |
| Arena Maths | Arena (LMArena) | 1477† | 1483† |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 53.0% | 43.2% |
| LiveBench 2.0 Data Analysis | LiveBench | 66.0%† | 79.3%† |
| LiveBench Research | LiveBench | 81.8%† | 60.4%† |
| LiveBench 1.0 Data Analysis | LiveBench | 66.0% | 79.3% |
| AA-Omniscience | Artificial Analysis | 28.8 | 0.1 |
| MMLU-Pro | Vals AI | 89.6%† | 86.7%† |
| LegalBench | Vals AI | 83.6%† | 85.1% |
| CorpFin | Vals AI | 66.7%† | 65.3%† |
| TaxEval | Vals AI | 75.6%† | 76.2%† |
| Excel Modeling Benchmark | Vals AI | 69.4%† | 66.2% |
| MedCode | Vals AI | 53.2%† | 43.4%† |
| MedScribe | Vals AI | 85.8%† | 82.9%† |
| Public Benefits Bench v1.1 | Vals AI | 68.1%† | 62.4% |
| DrugDiscoveryBench | Scale AI SEAL | 46.8% | – |
| EBR-bench | Epoch AI Benchmarking Hub | 28.6% | – |
| GDP.pdf | Epoch AI Benchmarking Hub | 24.0% | 24.7%† |
| AA-Omniscience: accuracy | Artificial Analysis | 48.8% | 46.8% |
| AA-Omniscience: non-hallucination | Artificial Analysis | 60.8% | 12.1% |
| Instruction following | |||
| LiveBench Language | LiveBench | 79.7% | 82.9% |
| LiveBench Instruction Following | LiveBench | 72.0% | 64.6% |
| IFBench | Artificial Analysis | 62.2% | 71.2% |
| Arena Instruction Following | Arena (LMArena) | 1478† | 1464† |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1474† | 1467† |
| EQ-Bench 4 | EQ-Bench | 1267† | 1224† |
| Arena Creative Writing | Arena (LMArena) | 1463† | 1429† |
| Arena Multi-turn | Arena (LMArena) | 1495† | 1472† |
| Arena Longer Queries | Arena (LMArena) | 1499† | 1468† |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1289† | 1271† |
| LiveBench Design | LiveBench | 78.8%† | 76.0%† |
| MMMU-Pro | Artificial Analysis | – | 80.7% |
| MMMU-Pro (Vals) | Vals AI | 86.6%† | 86.5% |
| MortgageTax | Vals AI | 69.9%† | 67.3%† |
| SAGE | Vals AI | 54.8%† | 47.0% |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 14.5%† | 30.8%† |
| Long context | |||
| AA-LCR | Artificial Analysis | 77.7% | 83.0% |
| Arena Document | Arena (LMArena) | 1475† | 1477† |
| MLCR | Artificial Analysis | 19.4% | 27.2% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 159.0† | 160.1† |
| LiveBench 2.0 | LiveBench | 76.8%† | 75.0%† |
| LiveBench 1.0 | LiveBench | 76.2% | 77.9% |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 41.8 | 42.1 |
| Vals RSI Index | Vals AI | 24.0% | – |
| Vals Multimodal Index | Vals AI | 70.9%† | 65.1%† |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | 18.2% | – |
Data as of 2026-10-25. Best configuration of each model; every score links to its source on the model pages.
What changed
- Claude Opus 4.8 (max): 1 new result (AA-Briefcase v1.1).
- Claude Opus 4.8 (max): 1 new result (Vals RSI Index).
- GPT-5.6 Terra (max): 1 new result (Harvey's Legal Agent Benchmark v1).
- Claude Opus 4.8 (max): 1 score revision.
- GPT-5.6 Terra (max): 1 score revision.
- Claude Opus 4.8 (max): 1 new result (ForecastBench).
- Claude Opus 4.8 (max): BenchLeader Index 62.5 → 63.4.
- GPT-5.6 Terra (max): BenchLeader Index 63.8 → 64.4.
- Claude Opus 4.8 (max): 5 new results (AutomationBench, GDP.pdf (AA), MLCR…).
- GPT-5.6 Terra (max): 5 new results (AutomationBench, GDP.pdf (AA), MLCR…).
Related comparisons
- Claude Opus 4.8 vs Claude Opus 5.5
- Claude Opus 4.8 vs Gemini 4 Argon
- Claude Opus 4.8 vs Muse Wonder Full
- GPT-5.6 Terra vs Claude Opus 5.5
- GPT-5.6 Terra vs Gemini 4 Argon
- GPT-5.6 Terra vs Muse Wonder Full
- GPT-5.6 Terra vs Claude Fable 5.1
- Claude Opus 4.7 vs GPT-5.6 Terra
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Claude Opus 4.8 vs GPT-5.6 Terra: questions
- Is Claude Opus 4.8 better than GPT-5.6 Terra?
- Claude Opus 4.8 (max) and GPT-5.6 Terra (max) are neck and neck on quality (63.3 vs 64.4, within the ±6.9 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 2026-10-25, so check the category scores for your use.
- Which is cheaper, Claude Opus 4.8 or GPT-5.6 Terra, and will that change?
- GPT-5.6 Terra is cheaper today: $4.50 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 better than GPT-5.6 Terra for coding?
- GPT-5.6 Terra scores higher in coding (65 vs 59 on the category index, where 50 is average). On SciCode, the top shared coding benchmark, it is 53.5% for Claude Opus 4.8 against 55.0% for GPT-5.6 Terra.
- Is Claude Opus 4.8 better than GPT-5.6 Terra for agents and tool use?
- GPT-5.6 Terra scores higher in agentic tasks (59 vs 57 on the category index, where 50 is average). On Terminal-Bench, the top shared agents & tools benchmark, it is 23.6% for Claude Opus 4.8 against 21.5% for GPT-5.6 Terra.
- Which is better for long documents?
- GPT-5.6 Terra scores higher on long-context benchmarks (67 vs 64). Both take about 1.1M tokens: Claude Opus 4.8 1,000,000, GPT-5.6 Terra 1,050,000.
- Which has lower latency?
- Claude Opus 4.8 (max) answers first: 29.3 s against 101.0 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.
- Which is faster, Claude Opus 4.8 or GPT-5.6 Terra?
- GPT-5.6 Terra streams faster: 126 against 63 output tokens per second.
- Do independent benchmarks agree on Claude Opus 4.8 vs GPT-5.6 Terra?
- Across the 40 benchmarks both have, Claude Opus 4.8 leads 15, GPT-5.6 Terra leads 25. Independent sources split: Claude Opus 4.8 leads on Arena Text; GPT-5.6 Terra leads on AA Intelligence Index v4.3.2.
- When was GPT-5.6 Terra released?
- GPT-5.6 Terra was released on 9 Jul 2026.