Claude Opus 5.5 vs GPT-5-Codex
Full profiles: Claude Opus 5.5 (Anthropic) and GPT-5-Codex (OpenAI): every benchmark, price and rank.
Updated · Data from 10 sources
As of 29 Oct 2026, Claude Opus 5.5 is ahead of GPT-5-Codex on quality (69.2 vs 58.6 on the BenchLeader Index), but GPT-5-Codex is 2.3× cheaper at $3.44 vs $8.00 per 1M tokens blended. Choose Claude Opus 5.5 for coding (80 vs 59), multimodal (71 vs 56), and long context (67 vs 61); choose GPT-5-Codex for fast first answers (10.9 s vs 648.2 s) and agents & tools (67 vs 63), where it also streams ~1.5× faster.
Key facts (as of 29 Oct 2026)
- Claude Opus 5.5 scores 69.2 on the BenchLeader Index and GPT-5-Codex scores 58.6 (BenchLeader, 29 Oct 2026).
- Claude Opus 5.5 costs $8.00 per 1M tokens blended (list price, 29 Oct 2026).
- GPT-5-Codex costs $3.44 per 1M tokens blended (list price, 29 Oct 2026).
- GPT-5-Codex streams ~1.5× faster than Claude Opus 5.5 (Artificial Analysis, 7-day median, 29 Oct 2026).
- Of the 5 benchmarks both models have, Claude Opus 5.5 leads 5 (BenchLeader, 29 Oct 2026).
Claude Opus 5.5 benchmarks max Anthropic
GPT-5-Codex benchmarks high 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 | Claude Opus 5.5 | 80 vs 59 category score | High |
| Agents & tool use | GPT-5-Codex | 67 vs 63 category score | Medium |
| Long documents | Claude Opus 5.5 | 67 vs 61 category score; AA-LCR v1.1 84.7% vs 71.7% | Medium |
| Images and other media | Claude Opus 5.5 | 71 vs 56 category score; MMMU-Pro 87.7% vs 73.8% | High |
| Lowest cost | GPT-5-Codex | $3.44 vs $8.00 per 1M blended; agentic coding session $0.115 vs $0.320 | – |
| Fastest first answer | GPT-5-Codex | 10.9 s (GPT-5-Codex (high)) vs 648.2 s (Claude Opus 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 Opus 5.5 (max) leads on quality: 69.2 vs 58.6.
- Claude Opus 5.5 (max) is stronger in coding, knowledge, long context, maths, multimodal, reasoning.
- GPT-5-Codex (high) is stronger in agents & tools, instruction following.
- GPT-5-Codex (high) is 2.3× cheaper ($3.44 vs $8.00 per 1M blended).
- GPT-5-Codex (high) streams 2.0× faster (195 vs 98 tokens per second).
At a glance
| Spec | Claude Opus 5.5 | GPT-5-Codex |
|---|---|---|
| Developer | Anthropic | OpenAI |
| Released | 22 Sept 2026 | 15 Sept 2025 |
| Context window (tokens) | 1,000,000 | 400,000 |
| Max output (tokens) | 128,000 (up to 300K on the Message Batches API (beta)) | – |
| Input types | text, image | – |
| Reasoning effort settings | low, medium, high, xhigh, max | – |
| Compared at | Claude Opus 5.5 (max) | GPT-5-Codex (high) |
| Open weights | No | No |
| Price in / out / cached input per 1M | $4.00 / $20.00 / $0.200 | $1.25 / $10.00 / $0.125 |
| API model ID | claude-opus-5-5 | – |
| Where to use it | Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry | – |
Limits, availability and announced prices from the vendors' own pages: docs.anthropic.com (checked 20 Oct 2026). Everything else is from the day's data.
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 5.5 (max) | GPT-5-Codex (high) | Leader |
|---|---|---|---|
| Reasoning | 72.9 | – | not comparable |
| Coding | 80.4 | 58.5 | Claude Opus 5.5 |
| Agents & tools | 63.1 | 66.8 | GPT-5-Codex |
| Maths | 74.0 | – | not comparable |
| Knowledge | 74.1 | – | not comparable |
| Instruction following | – | 72.4 | not comparable |
| Multimodal | 70.6 | 56.4 | Claude Opus 5.5 |
| Long context | 67.3 | 60.5 | Claude Opus 5.5 |
Where sources disagree
Across the 5 benchmarks both have, Claude Opus 5.5 leads 5, GPT-5-Codex leads 0. Every overall index that rates both puts Claude Opus 5.5 ahead.
| Overall index | Claude Opus 5.5 (max) | GPT-5-Codex (high) | Leader |
|---|---|---|---|
| BenchLeader Index | 69.2 | 58.6 | Claude Opus 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; suite cost is what Artificial Analysis paid to run its whole Intelligence Index.
| Measure | Claude Opus 5.5 (max) | GPT-5-Codex (high) |
|---|---|---|
| Blended price per 1M tokens | $8.00 | $3.44 |
| Input / output per 1M | $4.00 / $20.00 | $1.25 / $10.00 |
| Cached input per 1M | $0.200 | $0.125 |
| Price per task: chat reply | $0.0076 | $0.0035 |
| Price per task: summarize a 30-page report | $0.060 | $0.021 |
| Price per task: code edit | $0.054 | $0.022 |
| Price per task: agentic coding session | $0.320 | $0.115 |
| Price per task: structured extraction | $0.012 | $0.0045 |
| Cost to run a full benchmark suite | $10.04 | – |
| Output speed | 98 tok/s | 195 tok/s |
| Time to first answer | 648.2 s | 10.9 s |
| Time to a complete answer | 653.3 s | 13.4 s |
Each model is compared at its best-ranked reasoning-effort setting. Claude Opus 5.5 (max) scores 69.2 and takes 648 s to its first answer token; Claude Opus 5.5 (xhigh) scores 66.0 in 107 s, Claude Opus 5.5 (high) scores 65.6 in 34 s. GPT-5-Codex (high) scores 58.6 and takes 11 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
135 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 5.5 (max) | GPT-5-Codex (high) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 69.2 | 58.6 |
| Agents & tools score | BenchLeader | 63.1 | 66.8 |
| Coding score | BenchLeader | 80.4 | 58.5 |
| Knowledge score | BenchLeader | 74.1 | – |
| Long context score | BenchLeader | 67.3 | 60.5 |
| Maths score | BenchLeader | 74.0 | – |
| Multimodal score | BenchLeader | 70.6 | 56.4 |
| Reasoning score | BenchLeader | 72.9 | – |
| Instruction following score | BenchLeader | – | 72.4 |
| Blended price $/M | BenchLeader | $8.00 | $3.44 |
| Output speed | BenchLeader | 98 tok/s | 195 tok/s |
| Time to first answer | BenchLeader | 648.2 s | 10.9 s |
| Context window | BenchLeader | 1,000,000 tokens | 400,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 90.6% | – |
| Arena Hard Prompts | Arena (LMArena) | 1535† | – |
| LiveBench 2.0 Reasoning | LiveBench | 92.2%† | – |
| LiveBench 1.0 Reasoning | LiveBench | 92.2% | – |
| GPQA Diamond (AA) | Artificial Analysis | – | 83.7% |
| Humanity's Last Exam (AA) | Artificial Analysis | 61.4% | 27.9% |
| HLE-Diamond (AA) | Artificial Analysis | 67.9% | – |
| Kagi LLM Benchmark | Kagi LLM Benchmark | – | 70.3%† |
| ARC-AGI-1 | ARC Prize | 97.5% | – |
| ARC-AGI-2 | ARC Prize | 91.7% | – |
| CritPt | Artificial Analysis | 31.7% | 5.1% |
| CritPt v2 hard subset (AA) | Artificial Analysis | 61.0% | – |
| MysteryMechanism | Vals AI | 49.5%† | – |
| Chess Puzzles | Epoch AI Benchmarking Hub | 67.0% | – |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 71.0% | – |
| LMCA | Epoch AI Benchmarking Hub | 68.2% | – |
| DTBench | Epoch AI Benchmarking Hub | 98.9% | – |
| Coding | |||
| SciCode | SciCode | 66.9% | – |
| WeirdML | WeirdML | – | 54.5% |
| FrontierCode | Cognition | 54.6%† | – |
| Arena Coding | Arena (LMArena) | 1548† | – |
| Arena WebDev | Arena (LMArena) | 1809 | – |
| LiveBench 2.0 Coding | LiveBench | 96.3%† | – |
| LiveBench 1.0 Coding | LiveBench | 89.3% | – |
| LiveCodeBench | Vals AI | – | 84.7% |
Show all 135 benchmarks
| Metric | Publisher | Claude Opus 5.5 (max) | GPT-5-Codex (high) |
|---|---|---|---|
| Coding | |||
| IOI | Vals AI | 95.1%† | – |
| Code Migration | Vals AI | 66.7%† | – |
| Code Migration v1.1 | Vals AI | 81.0%† | – |
| ProgramBench | Vals AI | 18.5%† | – |
| SWE-bench Multimodal | Vals AI | 70.2%† | – |
| Vibe Code Bench 1-100 | Vals AI | 30.4%† | – |
| Vibe Code Bench v1.1 | Vals AI | 90.3%† | – |
| MirrorCode | Epoch AI Benchmarking Hub | 77.4% | – |
| CursorBench | Cursor | 57.8% | – |
| ALE-Bench | Epoch AI Benchmarking Hub | 2147.4† | – |
| FrontierSWE | Epoch AI Benchmarking Hub | 62.3% | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 59.6% | – |
| Startrise Frontend | Startrise | 71.1%† | – |
| FrontierCode 1.1 (AA) | Artificial Analysis | 52.2% | – |
| FrontierCode 2.0 (AA) | Artificial Analysis | 46.4% | – |
| LiveCodeBench (AA) | Artificial Analysis | – | 84.0% |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 64.8% | 44.3%† |
| Remote Labor Index | Scale AI / CAIS | 21.3%† | – |
| APEX-Agents | Mercor | 73.5% | – |
| Arena Agent | Arena (LMArena) | 20.8 | – |
| LiveBench Long Agentic | LiveBench | 72.8%† | – |
| LiveBench Browser | LiveBench | 88.4%† | – |
| LiveBench Agentic Coding | LiveBench | 71.7% | – |
| Terminal-Bench Hard | Artificial Analysis | – | 37.9% |
| τ²-Bench Telecom (AA) | Artificial Analysis | – | 86.8% |
| Terminal-Bench 2.1 (Vals) | Vals AI | 87.6%† | – |
| HiL-Bench | Scale AI SEAL | 59.0%† | – |
| GDPval-AA v2.1, normalised score | Artificial Analysis | 68.3% | – |
| GDPval-AA v2.1 | Artificial Analysis | 1866 | – |
| ITBench SRE (AA) | Artificial Analysis | 38.2% | – |
| Analyst Agent (AA) | Artificial Analysis | 56.3% | – |
| CUA-bench | Vals AI | 14.0% | – |
| CyberBench | Vals AI | 55.4%† | – |
| Finance Agent v2 | Vals AI | 58.6%† | – |
| Finance Agent v2.1 | Vals AI | 67.8%† | – |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | 9.2%† | – |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 3.8%† | – |
| Legal Research Bench | Vals AI | 50.5%† | – |
| Robo-bench | Vals AI | 14.6% | – |
| SREBench | Vals AI | 33.6%† | – |
| Tax Agent Bench | Vals AI | 45.6%† | – |
| Terminal-Bench 4.0 (Vals) | Vals AI | 65.2%† | – |
| Terminal-Bench Science | Vals AI | 47.1%† | – |
| Time Horizon Index: KSP | Vals AI | 91.3%† | – |
| Arena Search | Arena (LMArena) | 1211† | – |
| Vending-Bench 2 | Andon Labs | 9235.3† | – |
| AutomationBench | Artificial Analysis | 69.5% | – |
| GDP.pdf (AA) | Artificial Analysis | 26.2% | – |
| Harvey LAB-AA v1.1 | Harvey | 4.2% | – |
| AA-Briefcase v1.1 | Artificial Analysis | 1800 | – |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | 59.0% | – |
| CWE-Bench (AA) | Artificial Analysis | 58.3% | – |
| DeepSec-Bench (AA) | Artificial Analysis | 26.9% | – |
| CyberGym end-to-end (AA) | Artificial Analysis | 0.8% | – |
| FACTS Search | 86.5%† | – | |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 91.2% | – |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 95.0% | – |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 100.0% | – |
| ProofBench | Vals AI | 100.0% | – |
| LiveBench Mathematics | LiveBench | 97.1% | – |
| Arena Maths | Arena (LMArena) | 1512† | – |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 72.2% | – |
| LiveBench 2.0 Data Analysis | LiveBench | 80.3%† | – |
| LiveBench Research | LiveBench | 85.0%† | – |
| LiveBench 1.0 Data Analysis | LiveBench | 80.3% | – |
| AA-Omniscience v1.1 | Artificial Analysis | 48.1 | – |
| PRBench Legal | Scale AI SEAL | 47.6%† | – |
| BioMysteryBench | Vals AI | 79.3%† | – |
| ClaimBench | Vals AI | 100.0%† | – |
| Excel Modeling Benchmark | Vals AI | 75.9%† | – |
| Excel Modeling Benchmark v1.1 | Vals AI | 75.8%† | – |
| MedCode | Vals AI | 49.8%† | – |
| MedScribe | Vals AI | 91.4%† | – |
| Public Benefits Bench v1.1 | Vals AI | 70.6%† | – |
| EBR-bench | Epoch AI Benchmarking Hub | 71.4% | – |
| GDP.pdf | Epoch AI Benchmarking Hub | 30.6% | – |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | 65.5% | – |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | 49.7% | – |
| HealthBench Professional (AA) | OpenAI | 59.9% | – |
| Instruction following | |||
| LiveBench Language | LiveBench | 86.3% | – |
| LiveBench Instruction Following | LiveBench | 65.7% | – |
| IFBench | Artificial Analysis | – | 74.2% |
| Arena Instruction Following | Arena (LMArena) | 1521† | – |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1509† | – |
| EQ-Bench 4 | EQ-Bench | 1340† | – |
| Arena Creative Writing | Arena (LMArena) | 1520† | – |
| Arena Multi-turn | Arena (LMArena) | 1503† | – |
| Arena Longer Queries | Arena (LMArena) | 1524† | – |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1321† | – |
| LiveBench Design | LiveBench | 86.9%† | – |
| MMMU-Pro | Artificial Analysis | 87.7% | 73.8% |
| SAGE | Vals AI | 91.3%† | – |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 51.2%† | – |
| Code2Video-AA | Artificial Analysis | 1536 | – |
| AA-3D | Artificial Analysis | 1196 | – |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 84.7% | 71.7% |
| MLCR | Artificial Analysis | 27.8% | – |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 168.8† | – |
| LiveBench 2.0 | LiveBench | 85.1%† | – |
| LiveBench 1.0 | LiveBench | 83.2% | – |
| AA Intelligence Index v5.0 | Artificial Analysis | 58.4 | – |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 57.6 | 24.9 |
| AA Progress Index v1 | Artificial Analysis | 350 | 275.1 |
| Vals Index | Vals AI | 70.2† | – |
| Vals Finance Index | Vals AI | 65.6%† | – |
| Vals RSI Index | Vals AI | 37.3% | – |
| Vals Science Index | Vals AI | 58.6%† | – |
| Vals Router Index | Vals AI | 64.2%† | – |
| AA Cyber Index | Artificial Analysis | 28.7 | – |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | 5.8% | – |
Data as of 2026-10-29. Best configuration of each model; every score links to its source on the model pages.
What changed
- Claude Opus 5.5 (max): 5 new results (AA Intelligence Index v5.0, AA Progress Index v1, HLE-Diamond (AA)…); 3 score revisions; cost per task moved from $5.98/M to $10.04/M (68% up).
- GPT-5-Codex (high): 1 new result (AA Progress Index v1).
- Claude Opus 5.5 (max): 1 score revision.
- Claude Opus 5.5 (max): 2 new results (Robo-bench, Vals RSI Index).
- Claude Opus 5.5 (max): 2 score revisions.
- Claude Opus 5.5 (max): 2 score revisions.
- GPT-5-Codex (high): BenchLeader Index 59.0 → 58.5.
- Claude Opus 5.5 (max): 1 new result (FORTRESS).
- Claude Opus 5.5 (max): 1 score revision.
- Claude Opus 5.5 (max): 1 new result (Chess Puzzles); 1 score revision.
Related comparisons
- Claude Opus 5.5 vs GPT-6 Astra
- Claude Opus 5.5 vs GPT-6.1 Sol
- Claude Opus 5.5 vs Claude Fable 5.1
- GPT-5-Codex vs GPT-6 Astra
- GPT-5-Codex vs GPT-6.1 Sol
- GPT-5-Codex vs Claude Fable 5.1
- Claude Opus 5.5 vs GPT-5 Pro
- Claude Opus 5.5 vs Gemini 4 Argon
Rankings: Best LLM overall · Best LLM for coding · Best LLM for agents · Best cheap LLMs
Claude Opus 5.5 vs GPT-5-Codex: questions
- Is Claude Opus 5.5 better than GPT-5-Codex?
- Claude Opus 5.5 (max) leads on quality: 69.2 vs 58.6. The BenchLeader Index combines every independent quality benchmark; Claude Opus 5.5 (max) is ahead overall as of 29 Oct 2026, but check the category scores for your use.
- Claude Opus 5.5 vs GPT-5-Codex: which is cheaper, and will that change?
- GPT-5-Codex is cheaper today: $3.44 against $8.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is Claude Opus 5.5 or GPT-5-Codex better for coding?
- Claude Opus 5.5 scores higher in coding (80 vs 59 on the category index, where 50 is average).
- Which is faster, Claude Opus 5.5 or GPT-5-Codex?
- GPT-5-Codex streams ~1.5× faster (195 vs 98 output tokens per second, Artificial Analysis 7-day median).
- Is Claude Opus 5.5 or GPT-5-Codex better for agents and tool use?
- GPT-5-Codex scores higher in agentic tasks (67 vs 63 on the category index, where 50 is average).
More questions (4)
- Which is better for long documents?
- Claude Opus 5.5 scores higher on long-context benchmarks (67 vs 61). Claude Opus 5.5 accepts more context: 1,000,000 against 400,000 tokens for GPT-5-Codex.
- Which has lower latency?
- GPT-5-Codex answers first: 10.9 s against 648.2 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 Opus 5.5 vs GPT-5-Codex?
- Across the 5 benchmarks both have, Claude Opus 5.5 leads 5, GPT-5-Codex leads 0. Every overall index that rates both puts Claude Opus 5.5 ahead.
- When was Claude Opus 5.5 released?
- Claude Opus 5.5 was released on 22 Sept 2026.