Claude Sonnet 5 vs GPT-5.6 Terra
Full profiles: Claude Sonnet 5 (Anthropic) and GPT-5.6 Terra (OpenAI): every benchmark, price and rank.
Updated · Data from 10 sources
As of 16 Nov 2026, Claude Sonnet 5 and GPT-5.6 Terra are neck and neck on quality (59.0 vs 60.1 on the BenchLeader Index, inside the ±4.7 margin), and they cost about the same at $4.00 vs $4.50 per 1M tokens blended. Choose Claude Sonnet 5 for reasoning (65 vs 55); choose GPT-5.6 Terra for agents & tools (63 vs 53) and long context (64 vs 63).
Key facts (as of 16 Nov 2026)
- Claude Sonnet 5 scores 59.0 on the BenchLeader Index and GPT-5.6 Terra scores 60.1 (BenchLeader, 16 Nov 2026).
- Claude Sonnet 5 costs $4.00 per 1M tokens blended (list price, 16 Nov 2026).
- GPT-5.6 Terra costs $4.50 per 1M tokens blended (list price, 16 Nov 2026).
- Of the 28 benchmarks both models have, GPT-5.6 Terra leads 18 (BenchLeader, 16 Nov 2026).
- On Arena WebDev, Claude Sonnet 5 is rated 1539 and GPT-5.6 Terra 1520 (LMArena, as published by 16 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 | Either | 60 vs 59 category score; Arena WebDev 1539 vs 1520 | Either |
| Agents & tool use | GPT-5.6 Terra | 63 vs 53 category score; Arena Agent -1.8 vs 2 | High |
| Maths | Either | 65 vs 66 category score; Arena Maths 1472 vs 1478 | Either |
| Reasoning | Claude Sonnet 5 | 65 vs 55 category score; Arena Hard Prompts 1490 vs 1491 | High |
| Long documents | GPT-5.6 Terra | 64 vs 63 category score; AA-LCR v1.1 79.0% vs 76.7% | Medium |
| Images and other media | Either | 61 vs 61 category score; Arena Vision 1272 vs 1269 | Either |
| Chat quality (human preference) | Either | 65 vs 66 on Arena votes | Either |
| Lowest cost | Either | $4.00 vs $4.50 per 1M blended; agentic coding session $0.160 vs $0.168 | – |
| Fastest first answer | Either | 15.8 s (Claude Sonnet 5 (high)) vs 19.2 s (GPT-5.6 Terra (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 Sonnet 5 (high) and GPT-5.6 Terra (xhigh) are neck and neck on quality (59.0 vs 60.1, within the ±4.7 margin).
- Claude Sonnet 5 (high) is stronger in coding, reasoning.
- GPT-5.6 Terra (xhigh) is stronger in agents & tools, human preference, long context, maths, multimodal.
- They cost about the same ($4.00 per 1M blended).
At a glance
| Spec | Claude Sonnet 5 | GPT-5.6 Terra |
|---|---|---|
| Developer | Anthropic | OpenAI |
| Released | 30 Jun 2026 | 9 Jul 2026 |
| Context window (tokens) | 1,000,000 | 1,050,000 |
| Compared at | Claude Sonnet 5 (high) | GPT-5.6 Terra (xhigh) |
| Open weights | No | No |
| Price in / out / cached input per 1M | $2.00 / $10.00 / $0.200 | $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 Sonnet 5 (high) | GPT-5.6 Terra (xhigh) | Leader |
|---|---|---|---|
| Reasoning | 65.4 | 54.8 | Claude Sonnet 5 |
| Coding | 59.9 | 59.1 | level |
| Agents & tools | 52.9 | 63.1 | GPT-5.6 Terra |
| Maths | 65.3 | 65.7 | level |
| Knowledge | – | 53.0 | not comparable |
| Instruction following | – | 65.6 | not comparable |
| Human preference | 64.6 | 65.5 | level |
| Multimodal | 60.9 | 61.2 | level |
| Long context | 62.9 | 64.1 | GPT-5.6 Terra |
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: Knowledge, Instruction following.
Where sources disagree
Across the 28 benchmarks both have, Claude Sonnet 5 leads 10, GPT-5.6 Terra leads 18. Every overall index that rates both puts GPT-5.6 Terra ahead.
| Overall index | Claude Sonnet 5 (high) | GPT-5.6 Terra (xhigh) | Leader |
|---|---|---|---|
| BenchLeader Index | 59.0 | 60.1 | neck and neck |
| Arena Text | 1461 | 1468 | GPT-5.6 Terra |
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 Sonnet 5 (high) | GPT-5.6 Terra (xhigh) |
|---|---|---|
| Blended price per 1M tokens | $4.00 | $4.50 |
| Input / output per 1M | $2.00 / $10.00 | $2.00 / $12.00 |
| Cached input per 1M | $0.200 | $0.200 |
| Price per task: chat reply | $0.0038 | $0.0044 |
| Price per task: summarize a 30-page report | $0.030 | $0.031 |
| Price per task: code edit | $0.027 | $0.030 |
| Price per task: agentic coding session | $0.160 | $0.168 |
| Price per task: structured extraction | $0.0060 | $0.0064 |
| Agentic coding session, input mostly cached | $0.079 | $0.087 |
| Output speed | 99 tok/s | 97 tok/s |
| Time to first answer | 15.8 s | 19.2 s |
| Time to a complete answer, last measured 13 Nov 2026 | 40.7 s | 84.0 s |
Each model is compared at its best-ranked reasoning-effort setting. Claude Sonnet 5 (high) scores 59.0 and takes 16 s to its first answer token; Claude Sonnet 5 (max) scores 56.1 in 158 s, Claude Sonnet 5 (xhigh) scores 55.4 in 32 s. GPT-5.6 Terra (xhigh) scores 60.1 and takes 19 s to its first answer token; GPT-5.6 Terra (max) scores 59.5 in 73 s, GPT-5.6 Terra (high) scores 56.8 in 6 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
146 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 Sonnet 5 (high) | GPT-5.6 Terra (xhigh) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 59.0 | 60.1 |
| Agents & tools score | BenchLeader | 52.9 | 63.1 |
| Coding score | BenchLeader | 59.9 | 59.1 |
| Human preference score | BenchLeader | 64.6 | 65.5 |
| Long context score | BenchLeader | 62.9 | 64.1 |
| Maths score | BenchLeader | 65.3 | 65.7 |
| Multimodal score | BenchLeader | 60.9 | 61.2 |
| Reasoning score | BenchLeader | 65.4 | 54.8 |
| Instruction following score | BenchLeader | – | 65.6 |
| Knowledge score | BenchLeader | – | 53.0 |
| Blended price $/M | BenchLeader | $4.00 | $4.50 |
| Output speed | BenchLeader | 99 tok/s | 97 tok/s |
| Time to first answer | BenchLeader | 15.8 s | 19.2 s |
| Context window | BenchLeader | 1,000,000 tokens | 1,050,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 80.3%† | 93.3%† |
| SimpleBench | SimpleBench | 60.6%† | 48.9% |
| Arena Hard Prompts | Arena (LMArena) | 1490 | 1491 |
| LiveBench 2.0 Reasoning | LiveBench | – | 90.6%† |
| LiveBench 1.0 Reasoning | LiveBench | 88.7%† | 90.6%† |
| GPQA Diamond (AA) | Artificial Analysis | 91.1%† | 90.8% |
| Humanity's Last Exam (AA) | Artificial Analysis | 35.7% | 41.9% |
| HLE-Diamond (AA) | Artificial Analysis | – | 24.7% |
| GPQA Diamond (Vals) | Vals AI | 88.9%† | 90.9% |
| Kagi LLM Benchmark | Kagi LLM Benchmark | – | 51.3%† |
| ARC-AGI-1 | ARC Prize | – | 94.0% |
| ARC-AGI-2 | ARC Prize | – | 74.2% |
| ARC-AGI-3 | ARC Prize | – | 0.7% |
| CritPt | Artificial Analysis | 15.1% | 27.1% |
| CritPt v2 hard subset (AA) | Artificial Analysis | – | 50.5%† |
| Arena Expert | Arena (LMArena) | 1509 | 1507 |
| Chess Puzzles | Epoch AI Benchmarking Hub | 45.0%† | 47.0%† |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 16.0%† | 35.0%† |
| LMCA | Epoch AI Benchmarking Hub | 50.0% | 52.2% |
| DTBench | Epoch AI Benchmarking Hub | 84.5% | 92.5% |
| ForecastBench | ForecastBench | 63.1%† | – |
| BTF-3 | Epoch AI Benchmarking Hub | 14.2%† | – |
| Coding | |||
| SciCode | SciCode | 54.3% | 52.3% |
| WeirdML | WeirdML | 68.8% | 78.3%† |
| FrontierCode | Cognition | 42.7%† | 41.3%† |
Show all 146 benchmarks
| Metric | Publisher | Claude Sonnet 5 (high) | GPT-5.6 Terra (xhigh) |
|---|---|---|---|
| Coding | |||
| GSO-Bench | GSO-Bench | 37.3%† | – |
| Arena Coding | Arena (LMArena) | 1520 | 1517 |
| Arena WebDev | Arena (LMArena) | 1539 | 1520 |
| LiveBench 2.0 Coding | LiveBench | – | 84.2%† |
| LiveBench 1.0 Coding | LiveBench | 80.7%† | 78.3%† |
| LiveCodeBench | Vals AI | 82.4%† | 85.9% |
| IOI | Vals AI | 45.0%† | 87.6%† |
| SWE-bench (Vals) | Vals AI | 79.6%† | 95.4%† |
| Code Migration | Vals AI | 44.4%† | 47.8%† |
| ProgramBench | Vals AI | 0.0%† | 0.5%† |
| Vibe Code Bench 1-100 | Vals AI | 13.8%† | 14.8%† |
| Vibe Code Bench v1.1 | Vals AI | 81.3%† | 74.6%† |
| SWE-Bench Pro V2 | Scale AI SEAL | 93.2%† | 92.4% |
| SWE-Bench Pro V2 HARD | Scale AI SEAL | 88.2%† | 86.3% |
| MirrorCode | Epoch AI Benchmarking Hub | 63.3%† | 41.1%† |
| Surface Evolver Bench | Epoch AI Benchmarking Hub | 60.0%† | 83.8% |
| DeepSWE v1.1 | Epoch AI Benchmarking Hub | 48.2% | 60.2% |
| CursorBench | Cursor | 30.8% | 33.6% |
| ALE-Bench | Epoch AI Benchmarking Hub | 1463.1 | 1951.4† |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 5.0% | 10.1% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 80.5%† | 80.2% |
| Startrise Frontend | Startrise | 66.6%† | 66.9%† |
| GameBench 2 Overall | DuelLab GameBench 2 | 46.6† | 45.6† |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | 54.4† | 53.5 |
| Design Arena Website | Design Arena | 1268† | 1260† |
| FrontierCode 2.0 (AA) | Artificial Analysis | 17.8%† | 23.9%† |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 12.1%† | 21.5%† |
| APEX-Agents | Mercor | 54.5%† | 58.2%† |
| Arena Agent | Arena (LMArena) | 2 | -1.8 |
| LiveBench Long Agentic | LiveBench | – | 25.9%† |
| LiveBench Browser | LiveBench | – | 70.3%† |
| LiveBench Agentic Coding | LiveBench | 59.4%† | 55.0%† |
| Terminal-Bench Hard | Artificial Analysis | – | 62.9% |
| τ²-Bench Telecom (AA) | Artificial Analysis | – | 80.4% |
| Terminal-Bench 2.1 (Vals) | Vals AI | 74.5%† | 77.5%† |
| GDPval-AA v2.1, normalised score | Artificial Analysis | – | 47.4% |
| GDPval-AA v2.1 | Artificial Analysis | 1262 | 1456 |
| τ³-Banking (AA) | Artificial Analysis | 37.3%† | 29.7% |
| ITBench SRE (AA) | Artificial Analysis | – | 51.0%† |
| Analyst Agent (AA) | Artificial Analysis | 46.3%† | – |
| APEX-Agents (AA) | Artificial Analysis | – | 38.9%† |
| CyberBench | Vals AI | 61.9%† | 72.1%† |
| Finance Agent v2 | Vals AI | 53.9%† | 54.4%† |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 5.0%† | 0.8%† |
| Legal Research Bench | Vals AI | 41.8%† | 41.4%† |
| SkillsBench | Vals AI | 46.5%† | 58.9%† |
| SREBench | Vals AI | 0.4%† | 9.5%† |
| Tax Agent Bench | Vals AI | 29.4%† | 30.0%† |
| Terminal-Bench 4.0 (Vals) | Vals AI | 10.6%† | 22.7%† |
| Terminal-Bench Science | Vals AI | 5.7%† | 10.0%† |
| Arena Search | Arena (LMArena) | 1191† | – |
| Arena Agent Task Completion | Arena (LMArena) | -0.6 | -7.2 |
| BALROG | BALROG | – | 53.2%† |
| GBAEval | Epoch AI Benchmarking Hub | 65.3%† | – |
| Vending-Bench 2 | Andon Labs | 6377.7† | 7343.2† |
| AutomationBench | Artificial Analysis | 32.1% | 47.1% |
| GDP.pdf (AA) | Artificial Analysis | 9.4% | 24.6% |
| EnterpriseOps-Gym | Artificial Analysis | 44.7%† | 38.5%† |
| AA-Briefcase v1.1 | Artificial Analysis | 1177 | 1332 |
| Design Arena Agentic | Design Arena | 1222† | 1089† |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | – | 13.8%† |
| DeepSec-Bench (AA) | Artificial Analysis | – | 25.1%† |
| FACTS Search | – | 87.0%† | |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 65.6%† | 86.0%† |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 29.3%† | 70.7%† |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 80.0%† | 99.7%† |
| ProofBench | Vals AI | 77.0%† | 74.0% |
| LiveBench Mathematics | LiveBench | 92.9%† | 94.9%† |
| Arena Maths | Arena (LMArena) | 1472 | 1478 |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 33.7%† | 43.2%† |
| LiveBench 2.0 Data Analysis | LiveBench | – | 79.3%† |
| LiveBench Research | LiveBench | – | 79.7%† |
| LiveBench 1.0 Data Analysis | LiveBench | 71.7%† | 79.3%† |
| AA-Omniscience v1.1 | Artificial Analysis | – | -11.5 |
| MMLU-Pro | Vals AI | 87.5%† | 86.7% |
| LegalBench | Vals AI | 83.9%† | 85.1%† |
| CorpFin | Vals AI | 67.0%† | 65.3% |
| TaxEval | Vals AI | 75.6%† | 76.2% |
| Excel Modeling Benchmark | Vals AI | 66.3%† | 66.2%† |
| MedCode | Vals AI | 47.5%† | 43.4% |
| MedScribe | Vals AI | 76.0%† | 82.9% |
| Public Benefits Bench v1.1 | Vals AI | 66.0%† | 62.4%† |
| Financial Coaching Bench | Vals AI | – | 71.0%† |
| DrugDiscoveryBench | Scale AI SEAL | 50.0%† | – |
| GDP.pdf | Epoch AI Benchmarking Hub | – | 24.7%† |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | – | 41.3% |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | – | 9.9% |
| FACTS Parametric | 20.5%† | 57.6%† | |
| Instruction following | |||
| LiveBench Language | LiveBench | 75.0%† | 82.9%† |
| LiveBench Instruction Following | LiveBench | 63.9%† | 64.6%† |
| IFBench | Artificial Analysis | – | 66.3% |
| Arena Instruction Following | Arena (LMArena) | 1467 | 1463 |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1461 | 1468 |
| EQ-Bench 4 | EQ-Bench | 1221† | 1220† |
| EQ-Bench Creative Writing v3 | EQ-Bench | 1798† | 1856† |
| Arena Creative Writing | Arena (LMArena) | 1434 | 1430 |
| Arena Multi-turn | Arena (LMArena) | 1473 | 1475 |
| Arena Longer Queries | Arena (LMArena) | 1481 | 1469 |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1272 | 1269 |
| LiveBench Design | LiveBench | – | 70.9%† |
| MMMU-Pro | Artificial Analysis | 77.3%† | 79.5% |
| MMMU-Pro (Vals) | Vals AI | 83.0%† | 86.5%† |
| MortgageTax | Vals AI | 70.0%† | 67.3% |
| SAGE | Vals AI | – | 87.0%† |
| Humanity's Sixth Sense | Scale AI SEAL | 24.8%† | 22.7%† |
| GeoBench | GeoBench | 74.0% | – |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 24.9%† | 30.8%† |
| Furniture Assembly | Epoch AI Benchmarking Hub | – | 54.2%† |
| FACTS Multimodal | – | 39.9%† | |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 76.7% | 79.0% |
| Arena Document | Arena (LMArena) | 1469 | 1474 |
| MLCR | Artificial Analysis | 4.4% | 22.8% |
| FACTS Grounding | – | 60.5%† | |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 156.9† | 160.7† |
| LiveBench 2.0 | LiveBench | – | 67.2%† |
| LiveBench 1.0 | LiveBench | 76.0%† | 77.9%† |
| AA Intelligence Index v5.0.1 | Artificial Analysis | – | 32.3† |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 31.7 | 38.0 |
| AA Progress Index v1 | Artificial Analysis | 296.8 | 312 |
| Vals Multimodal Index | Vals AI | 68.8%† | 65.1% |
| Safety & honesty | |||
| DistressBench | Scale AI SEAL | 68.0%† | 76.4%† |
Data as of 2026-11-16. Best configuration of each model; every score links to its source on the model pages.
DuelLab GameBench 2, release gb2-20261111-7b598bb13de4 (https://benchmarks.duellab.org/releases/gb2-20261111-7b598bb13de4/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.
What changed
- Claude Sonnet 5 (high): 2 new results (GDPval-AA v2.1, AA-Briefcase v1.1).
- GPT-5.6 Terra (xhigh): 2 new results (GDPval-AA v2.1, AA-Briefcase v1.1).
- Claude Sonnet 5 (high): BenchLeader Index 59.5 → 59.0.
- Claude Sonnet 5 (high): 1 new result (GeoBench).
- Claude Sonnet 5 (high): 1 score revision.
- GPT-5.6 Terra (xhigh): 1 score revision.
- Claude Sonnet 5 (high): 1 score revision.
- GPT-5.6 Terra (xhigh): 1 score revision.
- GPT-5.6 Terra (xhigh): BenchLeader Index 61.2 → 60.4.
- Claude Sonnet 5 (high): 1 new result (AA Progress Index v1).
Related comparisons
- Claude Sonnet 5 vs Gemini 4 Carbon
- Claude Sonnet 5 vs Claude Opus 5.5
- Claude Sonnet 5 vs GPT-6 Astra
- GPT-5.6 Terra vs Gemini 4 Carbon
- Claude Sonnet 5 vs GPT-5.5
- Claude Sonnet 5 vs GPT-6.1 Sol
- Claude Sonnet 5 vs Gemini 4 Argon
- Claude Sonnet 5 vs Muse Spark
Rankings: Best LLM overall · Best LLM for agents · Best cheap LLMs
Claude Sonnet 5 vs GPT-5.6 Terra: questions
- Is Claude Sonnet 5 better than GPT-5.6 Terra?
- Claude Sonnet 5 (high) and GPT-5.6 Terra (xhigh) are neck and neck on quality (59.0 vs 60.1, within the ±4.7 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 16 Nov 2026, so check the category scores for your use.
- Claude Sonnet 5 vs GPT-5.6 Terra: which is cheaper, and will that change?
- Claude Sonnet 5 is cheaper today: $4.00 against $4.50 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is Claude Sonnet 5 or GPT-5.6 Terra better for coding?
- They are level in coding (60 vs 59 on the category index). On Arena WebDev, the top shared coding benchmark, it is 1539 for Claude Sonnet 5 against 1520 for GPT-5.6 Terra.
- Which is faster, Claude Sonnet 5 or GPT-5.6 Terra?
- Claude Sonnet 5 streams faster (99 vs 97 output tokens per second, Artificial Analysis 7-day median).
- Is Claude Sonnet 5 or GPT-5.6 Terra better for agents and tool use?
- GPT-5.6 Terra scores higher in agentic tasks (63 vs 53 on the category index, where 50 is average). On Arena Agent, the top shared agents & tools benchmark, it is 2 for Claude Sonnet 5 against -1.8 for GPT-5.6 Terra.
More questions (4)
- Which is better for long documents?
- GPT-5.6 Terra scores higher on long-context benchmarks (64 vs 63). Both take about 1.1M tokens: Claude Sonnet 5 1,000,000, GPT-5.6 Terra 1,050,000.
- Which has lower latency?
- Claude Sonnet 5 answers first: 15.8 s against 19.2 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 Sonnet 5 vs GPT-5.6 Terra?
- Across the 28 benchmarks both have, Claude Sonnet 5 leads 10, GPT-5.6 Terra leads 18. Every overall index that rates both puts GPT-5.6 Terra ahead.
- When was GPT-5.6 Terra released?
- GPT-5.6 Terra was released on 9 Jul 2026.