Claude Fable 5.1 vs GPT-5.6 Luna
Full profiles: Claude Fable 5.1 (Anthropic) and GPT-5.6 Luna (OpenAI): every benchmark, price and rank.
Updated · Data from 9 sources
As of 26 Oct 2026, Claude Fable 5.1 (max) leads on quality (70.5 vs 59.9 on the BenchLeader Index), but GPT-5.6 Luna (max) costs 44× less, so it is the better buy unless you need the extra quality. Claude Fable 5.1 leads in every category we measure both in.
Claude Fable 5.1 benchmarks max Anthropic
GPT-5.6 Luna 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 | Claude Fable 5.1 | 75 vs 56 category score; SciCode 63.1% vs 53.6% | High |
| Agents & tool use | Claude Fable 5.1 | 61 vs 49 category score; Terminal-Bench 57.9% vs 17.3% | High |
| Maths | Claude Fable 5.1 | 73 vs 65 category score; FrontierMath Tiers 1–3 90.2% vs 82.1% | High |
| Reasoning | Claude Fable 5.1 | 77 vs 63 category score; ARC-AGI-2 90.0% vs 59.6% | High |
| Knowledge | Claude Fable 5.1 | 78 vs 57 category score; SimpleQA Verified 70.8% vs 41.0% | High |
| Long documents | Either | 68 vs 67 category score; AA-LCR 85.3% vs 83.7% | Either |
| Images and other media | Claude Fable 5.1 | 67 vs 62 category score | Medium |
| Chat quality (human preference) | – | Only Claude Fable 5.1 has an Arena rating (70 category score) | Not comparable |
| Lowest cost | GPT-5.6 Luna | $0.450 vs $20.00 per 1M blended; agentic coding session $0.017 vs $0.800 | – |
| Fastest first answer | GPT-5.6 Luna | 120.3 s (GPT-5.6 Luna (max)) vs 180.5 s (Claude Fable 5.1 (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 Fable 5.1 (max) leads on quality: 70.5 vs 59.9.
- Claude Fable 5.1 (max) is stronger in agents & tools, coding, composite, human preference, knowledge, long context, maths, multimodal, reasoning.
- GPT-5.6 Luna (max) is 44× cheaper ($0.450 vs $20.00 per 1M blended).
- GPT-5.6 Luna (max) streams 1.8× faster (126 vs 70 tokens per second).
At a glance
| Spec | Claude Fable 5.1 | GPT-5.6 Luna |
|---|---|---|
| Developer | Anthropic | OpenAI |
| Released | 1 Sept 2026 | 9 Jul 2026 |
| Context window (tokens) | 1,000,000 | 1,050,000 |
| Compared at | Claude Fable 5.1 (max) | GPT-5.6 Luna (max) |
| Open weights | No | No |
| Price in / out / cached input per 1M | $10.00 / $50.00 / $0.250 | $0.200 / $1.20 / $0.020 |
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 Fable 5.1 (max) | GPT-5.6 Luna (max) | Leader |
|---|---|---|---|
| Reasoning | 76.8 | 62.9 | Claude Fable 5.1 |
| Coding | 74.9 | 56.0 | Claude Fable 5.1 |
| Agents & tools | 60.5 | 49.0 | Claude Fable 5.1 |
| Maths | 72.9 | 65.0 | Claude Fable 5.1 |
| Knowledge | 77.5 | 56.5 | Claude Fable 5.1 |
| Human preference | 70.0 | – | not comparable |
| Multimodal | 67.1 | 61.6 | Claude Fable 5.1 |
| Long context | 67.8 | 67.0 | level |
| Composite | 89.2 | 71.2 | Claude Fable 5.1 |
Where sources disagree
Across the 36 benchmarks both have, Claude Fable 5.1 leads 35, GPT-5.6 Luna leads 1. Every overall index that rates both puts Claude Fable 5.1 ahead.
| Overall index | Claude Fable 5.1 (max) | GPT-5.6 Luna (max) | Leader |
|---|---|---|---|
| BenchLeader Index | 70.5 | 59.9 | Claude Fable 5.1 |
| AA Intelligence Index v4.3.2 | 53.4 | 37.3 | Claude Fable 5.1 |
| Arena Text | 1503 | 1453† | Claude Fable 5.1 |
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 Fable 5.1 (max) | GPT-5.6 Luna (max) |
|---|---|---|
| Blended price per 1M tokens | $20.00 | $0.450 |
| Input / output per 1M | $10.00 / $50.00 | $0.200 / $1.20 |
| Cached input per 1M | $0.250 | $0.020 |
| Price per task: chat reply | $0.019 | $0.0004 |
| Price per task: summarize a 30-page report | $0.150 | $0.0031 |
| Price per task: code edit | $0.135 | $0.0030 |
| Price per task: agentic coding session | $0.800 | $0.017 |
| Price per task: structured extraction | $0.030 | $0.0006 |
| Cost to run a full benchmark suite | $7.63 | $0.178 |
| Output speed | 70 tok/s | 126 tok/s |
| Time to first answer | 180.5 s | 120.3 s |
| Time to a complete answer | 187.7 s | 124.3 s |
Per token is not per task: on Artificial Analysis' full suite Claude Fable 5.1 (max) cost $7.63 and GPT-5.6 Luna (max) $0.178, because GPT-5.6 Luna (max) used about 1.0× as many tokens for the same work.
Each model is compared at its best-ranked reasoning-effort setting. Claude Fable 5.1 (max) scores 70.5 and takes 181 s to its first answer token; Claude Fable 5.1 (xhigh) scores 69.5 in 91 s, Claude Fable 5.1 (high) scores 69.4 in 18 s. GPT-5.6 Luna (max) scores 59.9 and takes 120 s to its first answer token; GPT-5.6 Luna (xhigh) scores 59.7 in 42 s, GPT-5.6 Luna (high) scores 57.0 in 10 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 Fable 5.1 (max) | GPT-5.6 Luna (max) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 70.5 | 59.9 |
| Agents & tools score | BenchLeader | 60.5 | 49.0 |
| Coding score | BenchLeader | 74.9 | 56.0 |
| Composite score | BenchLeader | 89.2 | 71.2 |
| Human preference score | BenchLeader | 70.0 | – |
| Knowledge score | BenchLeader | 77.5 | 56.5 |
| Long context score | BenchLeader | 67.8 | 67.0 |
| Maths score | BenchLeader | 72.9 | 65.0 |
| Multimodal score | BenchLeader | 67.1 | 61.6 |
| Reasoning score | BenchLeader | 76.8 | 62.9 |
| Blended price $/M | BenchLeader | $20.00 | $0.450 |
| Output speed | BenchLeader | 70 tok/s | 126 tok/s |
| Time to first answer | BenchLeader | 180.5 s | 120.3 s |
| Context window | BenchLeader | 1,000,000 tokens | 1,050,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | – | 91.6% |
| Humanity's Last Exam | Scale AI / CAIS | 46.5%† | – |
| SimpleBench | SimpleBench | – | 46.8%† |
| Arena Hard Prompts | Arena (LMArena) | 1523 | 1475† |
| LiveBench 2.0 Reasoning | LiveBench | 91.7%† | – |
| LiveBench 1.0 Reasoning | LiveBench | 91.7% | 85.6% |
| GPQA Diamond (AA) | Artificial Analysis | 93.7% | 91.1% |
| Humanity's Last Exam (AA) | Artificial Analysis | 59.1% | 39.5% |
| GPQA Diamond (Vals) | Vals AI | 93.4%† | 91.7% |
| Kagi LLM Benchmark | Kagi LLM Benchmark | – | 49.1%† |
| ARC-AGI-1 | ARC Prize | 97.5% | 90.7% |
| ARC-AGI-2 | ARC Prize | 90.0% | 59.6% |
| ARC-AGI-3 | ARC Prize | – | 0.2% |
| CritPt | Artificial Analysis | 29.7% | 20.6% |
| MysteryMechanism | Vals AI | 47.8%† | 14.4% |
| Chess Puzzles | Epoch AI Benchmarking Hub | 48.0% | 43.0% |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 58.0% | 21.0% |
| LMCA | Epoch AI Benchmarking Hub | 65.5%† | 48.5% |
| DTBench | Epoch AI Benchmarking Hub | 97.6%† | 88.8% |
| Coding | |||
| SciCode | SciCode | 63.1% | 53.6% |
| WeirdML | WeirdML | 92.9% | 60.9%† |
| FrontierCode | Cognition | 50.9%† | 39.8%† |
| GSO-Bench | GSO-Bench | 88.2%† | – |
| Arena Coding | Arena (LMArena) | 1538 | 1502† |
| Arena WebDev | Arena (LMArena) | 1743 | 1519† |
Show all 143 benchmarks
| Metric | Publisher | Claude Fable 5.1 (max) | GPT-5.6 Luna (max) |
|---|---|---|---|
| Coding | |||
| LiveBench 2.0 Coding | LiveBench | 97.7%† | – |
| LiveBench 1.0 Coding | LiveBench | 86.4% | 82.9% |
| LiveCodeBench | Vals AI | 90.5%† | – |
| IOI | Vals AI | 90.8%† | 61.8% |
| SWE-bench (Vals) | Vals AI | – | 93.0% |
| Code Migration | Vals AI | 54.6%† | 44.5% |
| ProgramBench | Vals AI | 7.0%† | 0.0% |
| SWE-bench Multimodal | Vals AI | 61.3%† | – |
| Vibe Code Bench 1-100 | Vals AI | 28.0%† | 22.6% |
| Vibe Code Bench v1.1 | Vals AI | 90.3%† | 77.1% |
| SWE Atlas: Codebase QnA | Scale AI SEAL | 60.0%† | – |
| SWE Atlas: Refactoring | Scale AI SEAL | 56.7%† | – |
| SWE Atlas: Test Writing | Scale AI SEAL | 67.0%† | – |
| MirrorCode | Epoch AI Benchmarking Hub | 73.3%† | – |
| Surface Evolver Bench | Epoch AI Benchmarking Hub | – | 61.9%† |
| DeepSWE v1.1 | Epoch AI Benchmarking Hub | – | 67.2% |
| CursorBench | Cursor | 51.8% | 35.9% |
| ALE-Bench | Epoch AI Benchmarking Hub | 2143.2† | 1667.4 |
| FrontierSWE | Epoch AI Benchmarking Hub | 56.3% | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 52.0% | 11.6% |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | 91.4% | 80.9% |
| Startrise Frontend | Startrise | 77.1%† | 65.0%† |
| FrontierCode 1.1 (AA) | Artificial Analysis | 49.0% | – |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 57.9% | 17.3% |
| Remote Labor Index | Scale AI / CAIS | 17.9%† | – |
| APEX-Agents | Mercor | 68.6%† | 43.0%† |
| Arena Agent | Arena (LMArena) | 13.9 | – |
| LiveBench Long Agentic | LiveBench | 77.9%† | – |
| LiveBench Browser | LiveBench | 89.8%† | – |
| LiveBench Agentic Coding | LiveBench | 66.1% | 48.4% |
| Terminal-Bench 2.1 (Vals) | Vals AI | 85.0%† | 79.0% |
| MCP Atlas | Scale AI SEAL | 87.2%† | – |
| HiL-Bench | Scale AI SEAL | 61.5%† | – |
| GDPval-AA v2.1 | Artificial Analysis | 62.7% | – |
| τ³-Banking (AA) | Artificial Analysis | 47.2% | 31.1% |
| ITBench SRE (AA) | Artificial Analysis | 49.5% | 40.3% |
| Analyst Agent (AA) | Artificial Analysis | 57.5% | – |
| APEX-Agents (AA) | Artificial Analysis | – | 35.8% |
| CUA-bench | Vals AI | 13.2% | – |
| CyberBench | Vals AI | 70.4%† | 73.6% |
| Finance Agent v2 | Vals AI | 58.9%† | 55.0% |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 6.7%† | 1.3% |
| Legal Research Bench | Vals AI | 55.3%† | 36.5% |
| Robo-bench | Vals AI | 5.8% | – |
| SkillsBench | Vals AI | 61.5%† | 60.5% |
| SREBench | Vals AI | 22.9%† | 1.5% |
| Tax Agent Bench | Vals AI | 49.2%† | 24.8% |
| Terminal-Bench 4.0 (Vals) | Vals AI | 58.1%† | 11.6% |
| Terminal-Bench Science | Vals AI | 40.0%† | 0.0% |
| Time Horizon Index: KSP | Vals AI | 63.3%† | – |
| Arena Search | Arena (LMArena) | 1233† | – |
| BALROG | BALROG | – | 45.6% |
| Vending-Bench 2 | Andon Labs | 5421.6† | 4094.7† |
| AutomationBench | Artificial Analysis | 59.4% | 50.2% |
| GDP.pdf (AA) | Artificial Analysis | 26.2% | 24.0% |
| Harvey LAB-AA v1.1 | Harvey | 6.4% | – |
| EnterpriseOps-Gym | Artificial Analysis | – | 40.8% |
| AA-Briefcase v1.1 | Artificial Analysis | 1671 | 1339 |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | 43.3% | 3.3% |
| CWE-Bench (AA) | Artificial Analysis | 50.8% | – |
| DeepSec-Bench (AA) | Artificial Analysis | 23.5% | – |
| CyberGym end-to-end (AA) | Artificial Analysis | 0.8% | – |
| FACTS Search | – | 79.3%† | |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 90.2% | 82.1% |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 87.8% | 61.0% |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 100.0% | 98.3% |
| ProofBench | Vals AI | 100.0% | 60.0% |
| LiveBench Mathematics | LiveBench | 97.0% | 87.2% |
| Arena Maths | Arena (LMArena) | 1529 | 1478† |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 70.8% | 41.0% |
| LiveBench 2.0 Data Analysis | LiveBench | 80.3%† | – |
| LiveBench Research | LiveBench | 82.0%† | – |
| LiveBench 1.0 Data Analysis | LiveBench | 80.3% | 78.0% |
| AA-Omniscience | Artificial Analysis | 43.5 | -10.3 |
| MMLU-Pro | Vals AI | 92.4%† | 86.0% |
| LegalBench | Vals AI | 88.5%† | 84.0% |
| CorpFin | Vals AI | – | 64.2% |
| TaxEval | Vals AI | 76.0%† | 76.2% |
| PRBench Finance | Scale AI SEAL | 50.8%† | – |
| PRBench Legal | Scale AI SEAL | 51.6%† | – |
| BioMysteryBench | Vals AI | – | 61.5% |
| Excel Modeling Benchmark | Vals AI | 76.7%† | 67.1% |
| Excel Modeling Benchmark v1.1 | Vals AI | 76.7%† | – |
| MedCode | Vals AI | 53.5%† | 42.4% |
| MedScribe | Vals AI | 91.3%† | 84.4% |
| Public Benefits Bench v1.1 | Vals AI | 74.9%† | 61.2% |
| EBR-bench | Epoch AI Benchmarking Hub | 57.1% | – |
| GDP.pdf | Epoch AI Benchmarking Hub | 27.6% | 22.7%† |
| AA-Omniscience: accuracy | Artificial Analysis | 67.2% | 42.7% |
| AA-Omniscience: non-hallucination | Artificial Analysis | 27.4% | 7.4% |
| Instruction following | |||
| LiveBench Language | LiveBench | 89.5% | 72.6% |
| LiveBench Instruction Following | LiveBench | 73.0% | 60.1% |
| Arena Instruction Following | Arena (LMArena) | 1501 | 1444† |
| Human preference | |||
| Arena Text | Arena (LMArena) | 1503 | 1453† |
| EQ-Bench 4 | EQ-Bench | – | 1149† |
| Arena Creative Writing | Arena (LMArena) | 1487 | 1410† |
| Arena Multi-turn | Arena (LMArena) | 1488 | 1459† |
| Arena Longer Queries | Arena (LMArena) | 1511 | 1452† |
| Multimodal | |||
| Arena Vision | Arena (LMArena) | 1320 | 1257† |
| LiveBench Design | LiveBench | 83.4%† | – |
| MMMU-Pro | Artificial Analysis | – | 78.5% |
| MMMU-Pro (Vals) | Vals AI | 90.6%† | 85.0% |
| MortgageTax | Vals AI | 70.8%† | 67.3% |
| SAGE | Vals AI | 48.5%† | 44.2% |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 41.9%† | 22.6%† |
| AA-3D | Artificial Analysis | 1135 | – |
| Long context | |||
| AA-LCR | Artificial Analysis | 85.3% | 83.7% |
| Arena Document | Arena (LMArena) | 1493 | 1462† |
| MLCR | Artificial Analysis | 15.6% | 20.0% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 166.5† | 156.5† |
| LiveBench 2.0 | LiveBench | 85.7%† | – |
| LiveBench 1.0 | LiveBench | 83.4% | 73.6% |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 53.4 | 37.3 |
| Vals RSI Index | Vals AI | 36.1% | – |
| Vals Science Index | Vals AI | – | 25.3% |
| Vals Multimodal Index | Vals AI | – | 69.1% |
| AA Cyber Index | Artificial Analysis | 25.0 | – |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | 13.7%† | – |
Data as of 2026-10-26. Best configuration of each model; every score links to its source on the model pages.
What changed
- GPT-5.6 Luna (max): 1 new result (AA-Briefcase v1.1).
- Claude Fable 5.1 (max): 2 new results (Robo-bench, Vals RSI Index).
- GPT-5.6 Luna (max): 2 new results (Harvey's Legal Agent Benchmark v1, Vals Science Index).
- Claude Fable 5.1 (max): 1 score revision.
- GPT-5.6 Luna (max): 1 score revision.
- Claude Fable 5.1 (max): 1 score revision.
- Claude Fable 5.1 (max): BenchLeader Index 70.0 → 70.6.
- GPT-5.6 Luna (max): BenchLeader Index 58.8 → 59.9.
- GPT-5.6 Luna (max): 5 new results (AutomationBench, GDP.pdf (AA), MLCR…).
- GPT-5.6 Luna (max): 1 new result (SREBench); 1 score revision.
Related comparisons
- Claude Fable 5.1 vs Claude Opus 5.5
- Claude Fable 5.1 vs Gemini 4 Argon
- Claude Fable 5.1 vs Muse Wonder Full
- GPT-5.6 Luna vs Claude Opus 5.5
- GPT-5.6 Luna vs Gemini 4 Argon
- GPT-5.6 Luna vs Muse Wonder Full
- Claude Opus 5 vs GPT-5.6 Luna
- Claude Fable 5.1 vs GPT-5.5
Rankings: Best LLM overall · Best LLM for coding · Best LLM for agents · Best cheap LLMs
Claude Fable 5.1 vs GPT-5.6 Luna: questions
- Is Claude Fable 5.1 better than GPT-5.6 Luna?
- Claude Fable 5.1 (max) leads on quality: 70.5 vs 59.9. The BenchLeader Index combines every independent quality benchmark; Claude Fable 5.1 (max) is ahead overall as of 2026-10-26, but check the category scores for your use.
- Which is cheaper, Claude Fable 5.1 or GPT-5.6 Luna, and will that change?
- GPT-5.6 Luna is cheaper today: $0.450 against $20.00 per million tokens, blended at three input tokens per output token. No price change has been announced for either.
- Is Claude Fable 5.1 better than GPT-5.6 Luna for coding?
- Claude Fable 5.1 scores higher in coding (75 vs 56 on the category index, where 50 is average). On SciCode, the top shared coding benchmark, it is 63.1% for Claude Fable 5.1 against 53.6% for GPT-5.6 Luna.
- Is Claude Fable 5.1 better than GPT-5.6 Luna for agents and tool use?
- Claude Fable 5.1 scores higher in agentic tasks (61 vs 49 on the category index, where 50 is average). On Terminal-Bench, the top shared agents & tools benchmark, it is 57.9% for Claude Fable 5.1 against 17.3% for GPT-5.6 Luna.
- Which is better for long documents?
- They are level on long-context benchmarks (68 vs 67). Both take about 1.1M tokens: Claude Fable 5.1 1,000,000, GPT-5.6 Luna 1,050,000.
- Which has lower latency?
- GPT-5.6 Luna (max) answers first: 120.3 s against 180.5 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.
- Which is faster, Claude Fable 5.1 or GPT-5.6 Luna?
- GPT-5.6 Luna streams faster: 126 against 70 output tokens per second.
- Do independent benchmarks agree on Claude Fable 5.1 vs GPT-5.6 Luna?
- Across the 36 benchmarks both have, Claude Fable 5.1 leads 35, GPT-5.6 Luna leads 1. Every overall index that rates both puts Claude Fable 5.1 ahead.
- When was Claude Fable 5.1 released?
- Claude Fable 5.1 was released on 1 Sept 2026.