Claude Sonnet 5.5 vs GPT-5-Codex
Full profiles: Claude Sonnet 5.5 (Anthropic) and GPT-5-Codex (OpenAI): every benchmark, price and rank.
Updated · Data from 12 sources
As of 16 Nov 2026, Claude Sonnet 5.5 is ahead of GPT-5-Codex on quality (64.2 vs 58.6 on the BenchLeader Index), but GPT-5-Codex is 1.2× cheaper at $3.44 vs $4.00 per 1M tokens blended. Claude Sonnet 5.5 leads in every category we measure both in.
Key facts (as of 16 Nov 2026)
- Claude Sonnet 5.5 scores 64.2 on the BenchLeader Index and GPT-5-Codex scores 58.6 (BenchLeader, 16 Nov 2026).
- Claude Sonnet 5.5 costs $4.00 per 1M tokens blended (list price, 16 Nov 2026).
- GPT-5-Codex costs $3.44 per 1M tokens blended (list price, 16 Nov 2026).
- Of the 4 benchmarks both models have, Claude Sonnet 5.5 leads 4 (BenchLeader, 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 | Claude Sonnet 5.5 | 70 vs 59 category score | High |
| Agents & tool use | Claude Sonnet 5.5 | 71 vs 67 category score | Medium |
| Long documents | Claude Sonnet 5.5 | 66 vs 60 category score; AA-LCR v1.1 82.7% vs 71.7% | Medium |
| Lowest cost | GPT-5-Codex | $3.44 vs $4.00 per 1M blended; agentic coding session $0.115 vs $0.160 | – |
| Fastest first answer | GPT-5-Codex | 9.5 s (GPT-5-Codex (high)) vs 290.5 s (Claude Sonnet 5.5 (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 Sonnet 5.5 (max) leads on quality: 64.2 vs 58.6.
- Claude Sonnet 5.5 (max) is stronger in agents & tools, coding, long context.
- GPT-5-Codex (high) is 1.2× cheaper ($3.44 vs $4.00 per 1M blended).
At a glance
| Spec | Claude Sonnet 5.5 | GPT-5-Codex |
|---|---|---|
| Developer | Anthropic | OpenAI |
| Released | 28 Sep 2026 | 15 Sep 2025 |
| Context window (tokens) | 1,000,000 | 272,000 |
| Compared at | Claude Sonnet 5.5 (max) | GPT-5-Codex (high) |
| Open weights | No | No |
| Price in / out / cached input per 1M | $2.00 / $10.00 / $0.100 | $1.25 / $10.00 / $0.125 |
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 5.5 (max) | GPT-5-Codex (high) | Leader |
|---|---|---|---|
| Reasoning | 67.8 | – | not comparable |
| Coding | 70.2 | 58.6 | Claude Sonnet 5.5 |
| Agents & tools | 70.5 | 66.8 | Claude Sonnet 5.5 |
| Maths | 71.3 | – | not comparable |
| Knowledge | 57.4 | – | not comparable |
| Instruction following | – | 72.5 | not comparable |
| Multimodal | – | 56.1 | not comparable |
| Long context | 66.0 | 60.3 | Claude Sonnet 5.5 |
| Composite | 63.0 | – | 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: Reasoning, Maths, Knowledge, Instruction following, Multimodal, Human preference.
Where sources disagree
Across the 4 benchmarks both have, Claude Sonnet 5.5 leads 4, GPT-5-Codex leads 0. Every overall index that rates both puts Claude Sonnet 5.5 ahead.
| Overall index | Claude Sonnet 5.5 (max) | GPT-5-Codex (high) | Leader |
|---|---|---|---|
| BenchLeader Index | 64.2 | 58.6 | Claude Sonnet 5.5 |
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 5.5 (max) | GPT-5-Codex (high) |
|---|---|---|
| Blended price per 1M tokens | $4.00 | $3.44 |
| Input / output per 1M | $2.00 / $10.00 | $1.25 / $10.00 |
| Cached input per 1M | $0.100 | $0.125 |
| Price per task: chat reply | $0.0038 | $0.0035 |
| Price per task: summarize a 30-page report | $0.030 | $0.021 |
| Price per task: code edit | $0.027 | $0.022 |
| Price per task: agentic coding session | $0.160 | $0.115 |
| Price per task: structured extraction | $0.0060 | $0.0045 |
| Agentic coding session, input mostly cached | $0.074 | $0.064 |
| Cost per benchmark task (AA Index) | $10.02 | – |
| Output tokens per task (AA Index) | 285k | – |
| Of which reasoning | 206k | – |
| Time per task (AA Index) | 31 min | – |
| Output speed | 152 tok/s | 120 tok/s |
| Time to first answer | 290.5 s | 9.5 s |
| Time to a complete answer, last measured 13 Nov 2026 | 487.7 s | 17.8 s |
Each model is compared at its best-ranked reasoning-effort setting. Claude Sonnet 5.5 (max) scores 64.2 and takes 291 s to its first answer token; Claude Sonnet 5.5 (xhigh) scores 62.5 in 27 s, Claude Sonnet 5.5 (high) scores 59.5 in 9 s. GPT-5-Codex (high) scores 58.6 and takes 9 s to its first answer token. 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
151 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 5.5 (max) | GPT-5-Codex (high) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 64.2 | 58.6 |
| Agents & tools score | BenchLeader | 70.5 | 66.8 |
| Coding score | BenchLeader | 70.2 | 58.6 |
| Composite score | BenchLeader | 63.0 | – |
| Knowledge score | BenchLeader | 57.4 | – |
| Long context score | BenchLeader | 66.0 | 60.3 |
| Maths score | BenchLeader | 71.3 | – |
| Reasoning score | BenchLeader | 67.8 | – |
| Instruction following score | BenchLeader | – | 72.5 |
| Multimodal score | BenchLeader | – | 56.1 |
| Blended price $/M | BenchLeader | $4.00 | $3.44 |
| Output speed | BenchLeader | 152 tok/s | 120 tok/s |
| Time to first answer | BenchLeader | 290.5 s | 9.5 s |
| Context window | BenchLeader | 1,000,000 tokens | 272,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 95.6% | – |
| Humanity's Last Exam | Scale AI / CAIS | 46.1%† | – |
| SimpleBench | SimpleBench | 75.9%† | – |
| Arena Hard Prompts | Arena (LMArena) | 1499† | – |
| LiveBench 2.0 Reasoning | LiveBench | 91.6% | – |
| LiveBench 1.0 Reasoning | LiveBench | 91.6% | – |
| GPQA Diamond (AA) | Artificial Analysis | – | 83.7% |
| Humanity's Last Exam (AA) | Artificial Analysis | 55.0% | 27.9% |
| HLE-Diamond (AA) | Artificial Analysis | 53.3% | – |
| Kagi LLM Benchmark | Kagi LLM Benchmark | – | 70.3%† |
| ARC-AGI-1 | ARC Prize | 96.5% | – |
| ARC-AGI-2 | ARC Prize | 83.9%† | – |
| ARC-AGI-3 | ARC Prize | 72.4%† | – |
| CritPt | Artificial Analysis | 31.4% | 5.1% |
| CritPt v2 hard subset (AA) | Artificial Analysis | 59.0% | – |
| MysteryMechanism | Vals AI | 49.1% | – |
| HLE-Diamond (Scale) | Scale AI SEAL | 37.5%† | – |
| Arena Expert | Arena (LMArena) | 1512† | – |
| Chess Puzzles | Epoch AI Benchmarking Hub | 59.0% | – |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 65.0% | – |
| LMCA | Epoch AI Benchmarking Hub | 68.3% | – |
| DTBench | Epoch AI Benchmarking Hub | 97.3% | – |
| Coding | |||
| SciCode | SciCode | 61.0% | – |
| WeirdML | WeirdML | – | 54.5% |
| FrontierCode | Cognition | 52.1%† | – |
Show all 151 benchmarks
| Metric | Publisher | Claude Sonnet 5.5 (max) | GPT-5-Codex (high) |
|---|---|---|---|
| Coding | |||
| Arena Coding | Arena (LMArena) | 1517† | – |
| Arena WebDev | Arena (LMArena) | 1772† | – |
| LiveBench 2.0 Coding | LiveBench | 77.2% | – |
| LiveBench 1.0 Coding | LiveBench | 91.4% | – |
| LiveCodeBench | Vals AI | – | 84.7% |
| IOI | Vals AI | 83.1% | – |
| Code Migration | Vals AI | 69.8% | – |
| Code Migration v1.1 | Vals AI | 86.9% | – |
| ProgramBench | Vals AI | 6.5% | – |
| SWE-bench Multimodal | Vals AI | 58.3% | – |
| Vibe Code Bench v1.1 | Vals AI | 92.4% | – |
| SWE Atlas v1.1 | Scale AI SEAL | 68.3%† | – |
| SWE Atlas v1.1: Codebase QnA | Scale AI SEAL | 74.2%† | – |
| SWE Atlas v1.1: Test Writing | Scale AI SEAL | 67.8%† | – |
| SWE Atlas v1.1: Refactoring | Scale AI SEAL | 62.9%† | – |
| SWE-Bench Pro V2 | Scale AI SEAL | 95.8%† | – |
| SWE-Bench Pro V2 HARD | Scale AI SEAL | 70.6%† | – |
| MirrorCode | Epoch AI Benchmarking Hub | 97.8% | – |
| CursorBench | Cursor | 55.5% | – |
| ALE-Bench | Epoch AI Benchmarking Hub | 1819.1† | – |
| FrontierSWE | Epoch AI Benchmarking Hub | 61.9% | – |
| WeirdML v3 | WeirdML | 19.9%† | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 63.6% | – |
| Startrise Frontend | Startrise | 75.9%† | – |
| GameBench 2 Overall | DuelLab GameBench 2 | 60.4† | – |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | 91.4† | – |
| Design Arena Website | Design Arena | 1309† | – |
| FrontierCode 1.1 (AA) | Artificial Analysis | 45.8% | – |
| FrontierCode 2.0 (AA) | Artificial Analysis | 41.7% | – |
| LiveCodeBench (AA) | Artificial Analysis | – | 84.0% |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 63.9% | 44.3%† |
| APEX-Agents | Mercor | 75.5% | – |
| Arena Agent | Arena (LMArena) | 16 | – |
| LiveBench Long Agentic | LiveBench | 63.4% | – |
| LiveBench Browser | LiveBench | 89.3% | – |
| LiveBench Agentic Coding | LiveBench | 56.3% | – |
| Terminal-Bench Hard | Artificial Analysis | – | 37.9% |
| τ²-Bench Telecom (AA) | Artificial Analysis | – | 86.8% |
| Terminal-Bench 2.1 (Vals) | Vals AI | 83.2%† | – |
| GDPval-AA v2.1, normalised score | Artificial Analysis | 66.9% | – |
| GDPval-AA v2.1 | Artificial Analysis | 1838 | – |
| Analyst Agent (AA) | Artificial Analysis | 57.5% | – |
| CyberBench | Vals AI | 59.6% | – |
| Finance Agent v2 | Vals AI | 58.1% | – |
| Finance Agent v2.1 | Vals AI | 67.3% | – |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | 10.4% | – |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 2.9% | – |
| Legal Research Bench | Vals AI | 48.1% | – |
| SREBench | Vals AI | 30.1% | – |
| Tax Agent Bench | Vals AI | 42.0% | – |
| Terminal-Bench 4.0 (Vals) | Vals AI | 64.1% | – |
| Terminal-Bench Science | Vals AI | 47.1% | – |
| Arena Agent Task Completion | Arena (LMArena) | 19.5 | – |
| Vending-Bench 2 | Andon Labs | 11165.3† | – |
| SpireClimb | Epoch AI Benchmarking Hub | 1.3% | – |
| AutomationBench | Artificial Analysis | 71.8% | – |
| GDP.pdf (AA) | Artificial Analysis | 25.8% | – |
| Harvey LAB-AA v1.1 | Harvey | 2.8% | – |
| AA-Briefcase v1.1 | Artificial Analysis | 1816 | – |
| GDPx | GDPx by SemiAnalysis | 8.2% | – |
| Design Arena Agentic | Design Arena | 1293† | 1050† |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | 53.3% | – |
| MKA (AA) | Artificial Analysis | 52.0% | – |
| FACTS Search | 86.8%† | – | |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 88.8% | – |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 80.5% | – |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 100.0% | – |
| ProofBench | Vals AI | 100.0% | – |
| LiveBench Mathematics | LiveBench | 96.1% | – |
| Arena Maths | Arena (LMArena) | 1486† | – |
| FrontierMath Erdős | Epoch AI Benchmarking Hub | 2.9% | – |
| AIME 2025 (AA) | Artificial Analysis | – | 98.7% |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 46.5% | – |
| LiveBench 2.0 Data Analysis | LiveBench | 59.5% | – |
| LiveBench Research | LiveBench | 88.0% | – |
| LiveBench 1.0 Data Analysis | LiveBench | 59.5% | – |
| AA-Omniscience v1.1 | Artificial Analysis | 30.5 | – |
| BioMysteryBench | Vals AI | 81.1% | – |
| ClaimBench | Vals AI | 100.0% | – |
| Excel Modeling Benchmark | Vals AI | 75.7% | – |
| Excel Modeling Benchmark v1.1 | Vals AI | 75.9% | – |
| MedCode | Vals AI | 52.9% | – |
| MedScribe | Vals AI | 91.1% | – |
| Public Benefits Bench v1.1 | Vals AI | 67.2% | – |
| Access to Justice Benchmark | Vals AI | 74.9% | – |
| Financial Coaching Bench | Vals AI | 74.9% | – |
| Immigration Bench | Vals AI | 69.0% | – |
| GDP.pdf | Epoch AI Benchmarking Hub | 31.4% | – |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | 51.5% | – |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | 56.7% | – |
| HealthBench Professional (AA) | OpenAI | 63.0% | – |
| FACTS Parametric | 35.1%† | – | |
| Instruction following | |||
| LiveBench Language | LiveBench | 78.0% | – |
| LiveBench Instruction Following | LiveBench | 56.8% | – |
| IFBench | Artificial Analysis | – | 74.2% |
| Arena Instruction Following | Arena (LMArena) | 1482† | – |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1473† | – |
| Arena Creative Writing | Arena (LMArena) | 1464† | – |
| Arena Multi-turn | Arena (LMArena) | 1469† | – |
| Arena Longer Queries | Arena (LMArena) | 1491† | – |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1284† | – |
| LiveBench Design | LiveBench | 81.3% | – |
| MMMU-Pro | Artificial Analysis | – | 73.8% |
| SAGE | Vals AI | 88.0% | – |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 39.3%† | – |
| Furniture Assembly | Epoch AI Benchmarking Hub | 75.0% | – |
| VIALS-AA | Handshake AI | 59.6% | – |
| Code2Video-AA | Artificial Analysis | 1443 | – |
| AA-3D | Artificial Analysis | 1204 | – |
| FACTS Multimodal | 50.4%† | – | |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 82.7% | 71.7% |
| MLCR | Artificial Analysis | 41.7% | – |
| FACTS Grounding | 62.5%† | – | |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 166.6† | – |
| LiveBench 2.0 | LiveBench | 77.0% | – |
| LiveBench 1.0 | LiveBench | 75.7% | – |
| AA Intelligence Index v5.0.1 | Artificial Analysis | 54.5 | – |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 56 | 24.9 |
| AA Progress Index v1 | Artificial Analysis | 350.5 | 275.1 |
| Vals Index | Vals AI | 70.8 | – |
| Vals Finance Index | Vals AI | 64.5% | – |
| Vals Science Index | Vals AI | 59.1% | – |
| Vals Router Index | Vals AI | 69.0%† | – |
| Vals Legal Index | Vals AI | 68.0% | – |
| Safety & honesty | |||
| DistressBench | Scale AI SEAL | 74.6%† | – |
| Metadata | |||
| AA Openness Index | Artificial Analysis | – | 5.6 |
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 5.5 (max): 36 new results (LiveBench 2.0, LiveBench Long Agentic, LiveBench 2.0 Reasoning…).
- GPT-5-Codex (high): 1 new result (AIME 2025 (AA)).
- Claude Sonnet 5.5 (max): BenchLeader Index 65.1 → 64.6.
- Claude Sonnet 5.5 (max): 1 new result (MirrorCode).
- Claude Sonnet 5.5 (max): 2 score revisions.
- Claude Sonnet 5.5 (max): 1 score revision.
- Claude Sonnet 5.5 (max): 1 score revision.
- Claude Sonnet 5.5 (max): 1 score revision.
- Claude Sonnet 5.5 (max): 1 new result (MirrorCode).
- Claude Sonnet 5.5 (max): 5 new results (AA Intelligence Index v5.0.1, AA Progress Index v1, HLE-Diamond (AA)…); 3 score revisions; cost per task moved from $5.46/M to $10.02/M (83% up).
Related comparisons
- Claude Sonnet 5.5 vs Claude Fable 5.2
- Claude Sonnet 5.5 vs Gemini 4 Carbon
- Claude Sonnet 5.5 vs Claude Opus 5.5
- GPT-5-Codex vs Claude Fable 5.2
- GPT-5-Codex vs Gemini 4 Carbon
- GPT-5-Codex vs Claude Opus 5.5
- Claude Sonnet 5.5 vs GPT-5 Pro
- Claude Sonnet 5.5 vs Claude Fable 5.1
Rankings: Best LLM overall · Best LLM for coding · Best LLM for agents · Best cheap LLMs
Claude Sonnet 5.5 vs GPT-5-Codex: questions
- Is Claude Sonnet 5.5 better than GPT-5-Codex?
- Claude Sonnet 5.5 (max) leads on quality: 64.2 vs 58.6. The BenchLeader Index combines every independent quality benchmark; Claude Sonnet 5.5 (max) is ahead overall as of 16 Nov 2026, but check the category scores for your use.
- Claude Sonnet 5.5 vs GPT-5-Codex: which is cheaper, and will that change?
- GPT-5-Codex is cheaper today: $3.44 against $4.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is Claude Sonnet 5.5 or GPT-5-Codex better for coding?
- Claude Sonnet 5.5 scores higher in coding (70 vs 59 on the category index, where 50 is average).
- Which is faster, Claude Sonnet 5.5 or GPT-5-Codex?
- Claude Sonnet 5.5 streams faster (152 vs 120 output tokens per second, Artificial Analysis 7-day median).
- Is Claude Sonnet 5.5 or GPT-5-Codex better for agents and tool use?
- Claude Sonnet 5.5 scores higher in agentic tasks (71 vs 67 on the category index, where 50 is average).
More questions (4)
- Which is better for long documents?
- Claude Sonnet 5.5 scores higher on long-context benchmarks (66 vs 60). Claude Sonnet 5.5 accepts more context: 1,000,000 against 272,000 tokens for GPT-5-Codex.
- Which has lower latency?
- GPT-5-Codex answers first: 9.5 s against 290.5 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 5.5 vs GPT-5-Codex?
- Across the 4 benchmarks both have, Claude Sonnet 5.5 leads 4, GPT-5-Codex leads 0. Every overall index that rates both puts Claude Sonnet 5.5 ahead.
- When was Claude Sonnet 5.5 released?
- Claude Sonnet 5.5 was released on 28 Sep 2026.