Claude Fable 5.2 vs GPT-6.2 Astra
Full profiles: Claude Fable 5.2 (Anthropic) and GPT-6.2 Astra (OpenAI): every benchmark, price and rank.
Updated · Data from 13 sources
As of 19 Nov 2026, Claude Fable 5.2 and GPT-6.2 Astra are neck and neck on quality (71.5 vs 68.9 on the BenchLeader Index, inside the ±4.4 margin), and they cost about the same at $16.00 vs $16.00 per 1M tokens blended. Claude Fable 5.2 leads in every category we measure both in. Claude Fable 5.2 streams ~2× faster.
Key facts (as of 19 Nov 2026)
- Claude Fable 5.2 scores 71.5 on the BenchLeader Index and GPT-6.2 Astra scores 68.9 (BenchLeader, 19 Nov 2026).
- Claude Fable 5.2 costs $16.00 per 1M tokens blended (list price, 19 Nov 2026).
- GPT-6.2 Astra costs $16.00 per 1M tokens blended (list price, 19 Nov 2026).
- Claude Fable 5.2 streams ~2× faster than GPT-6.2 Astra (Artificial Analysis, 7-day median, 19 Nov 2026).
- Of the 33 benchmarks both models have, Claude Fable 5.2 leads 17 (BenchLeader, 19 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 Fable 5.2 | 78 vs 76 category score; FrontierCode 2.0 (AA) 48.8% vs 45.4% | Medium |
| Agents & tool use | Claude Fable 5.2 | 75 vs 66 category score; Terminal-Bench 65.8% vs 62.1% | High |
| Maths | Claude Fable 5.2 | 74 vs 73 category score; FrontierMath Tiers 1–3 95.1% vs 92.6% | Medium |
| Reasoning | Either | 74 vs 73 category score; GPQA Diamond 93.9% vs 96.8% | Either |
| Knowledge | Claude Fable 5.2 | 85 vs 74 category score; SimpleQA Verified 94.0% vs 78.5% | High |
| Lowest cost | Either | $16.00 vs $16.00 per 1M blended; agentic coding session $0.640 vs $0.640 | – |
| Fastest first answer | Either | 221.3 s (Claude Fable 5.2 (max)) vs 202.7 s (GPT-6.2 Astra (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 5.2 (max) and GPT-6.2 Astra (max) are neck and neck on quality (71.5 vs 68.9, within the ±4.4 margin).
- Claude Fable 5.2 (max) is stronger in agents & tools, coding, knowledge, maths, reasoning.
- They cost about the same ($16.00 per 1M blended).
- Claude Fable 5.2 (max) streams 2.3× faster (110 vs 49 tokens per second).
At a glance
| Spec | Claude Fable 5.2 | GPT-6.2 Astra |
|---|---|---|
| Developer | Anthropic | OpenAI |
| Released | 9 Nov 2026 | 17 Nov 2026 |
| Context window (tokens) | 1,000,000 | 1,050,000 |
| Compared at | Claude Fable 5.2 (max) | GPT-6.2 Astra (max) |
| Open weights | No | No |
| Price in / out / cached input per 1M | $8.00 / $40.00 / $0.200 | $8.00 / $40.00 / $0.400 |
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 5.2 (max) | GPT-6.2 Astra (max) | Leader |
|---|---|---|---|
| Reasoning | 73.9 | 73.1 | level |
| Coding | 77.6 | 76.1 | Claude Fable 5.2 |
| Agents & tools | 75.2 | 66.4 | Claude Fable 5.2 |
| Maths | 74.0 | 72.5 | Claude Fable 5.2 |
| Knowledge | 84.5 | 74.2 | Claude Fable 5.2 |
| Multimodal | – | 71.7 | not comparable |
| Long context | 66.4 | – | not comparable |
| Composite | – | 69.5 | 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: Instruction following, Long context, Multimodal, Human preference.
Where sources disagree
Across the 33 benchmarks both have, Claude Fable 5.2 leads 17, GPT-6.2 Astra leads 15 and 1 is a tie. Every overall index that rates both puts Claude Fable 5.2 ahead.
| Overall index | Claude Fable 5.2 (max) | GPT-6.2 Astra (max) | Leader |
|---|---|---|---|
| BenchLeader Index | 71.5 | 68.9 | neck and neck |
| AA Intelligence Index v5.0.1 | 64.6 | 62.1 | Claude Fable 5.2 |
| Vals Index | 71.3† | 67.7 | Claude Fable 5.2 |
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 Fable 5.2 (max) | GPT-6.2 Astra (max) |
|---|---|---|
| Blended price per 1M tokens | $16.00 | $16.00 |
| Input / output per 1M | $8.00 / $40.00 | $8.00 / $40.00 |
| Cached input per 1M | $0.200 | $0.400 |
| Price per task: chat reply | $0.015 | $0.015 |
| Price per task: summarize a 30-page report | $0.120 | $0.120 |
| Price per task: code edit | $0.108 | $0.108 |
| Price per task: agentic coding session | $0.640 | $0.640 |
| Price per task: structured extraction | $0.024 | $0.024 |
| Agentic coding session, input mostly cached | $0.289 | $0.298 |
| Cost per benchmark task (AA Index) | $26.54 | $4.00 |
| Output tokens per task (AA Index) | 274k | 41k |
| Of which reasoning | 193k | 26k |
| Time per task (AA Index) | 42 min | 14 min |
| Output speed | 110 tok/s | 49 tok/s |
| Time to first answer | 221.3 s | 202.7 s |
| Time to a complete answer, last measured 13 Nov 2026 | 456.8 s | – |
Per token is not per task: an average task of Artificial Analysis' Intelligence Index cost $26.54 on Claude Fable 5.2 (max) and $4.00 on GPT-6.2 Astra (max), and Claude Fable 5.2 (max) wrote 6.7× as many output tokens for the same work, which a per-token price hides.
Each model is compared at its best-ranked reasoning-effort setting. Claude Fable 5.2 (max) scores 71.5 and takes 221 s to its first answer token; Claude Fable 5.2 (xhigh) scores 68.3 in 53 s, Claude Fable 5.2 (high) scores 67.3 in 24 s. GPT-6.2 Astra (max) scores 68.9 and takes 203 s to its first answer token; GPT-6.2 Astra (xhigh) scores 67.1 in 67 s, GPT-6.2 Astra (medium) scores 65.6 in 4 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
113 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 5.2 (max) | GPT-6.2 Astra (max) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 71.5 | 68.9 |
| Agents & tools score | BenchLeader | 75.2 | 66.4 |
| Coding score | BenchLeader | 77.6 | 76.1 |
| Knowledge score | BenchLeader | 84.5 | 74.2 |
| Long context score | BenchLeader | 66.4 | – |
| Maths score | BenchLeader | 74.0 | 72.5 |
| Reasoning score | BenchLeader | 73.9 | 73.1 |
| Composite score | BenchLeader | – | 69.5 |
| Multimodal score | BenchLeader | – | 71.7 |
| Blended price $/M | BenchLeader | $16.00 | $16.00 |
| Output speed | BenchLeader | 110 tok/s | 49 tok/s |
| Time to first answer | BenchLeader | 221.3 s | 202.7 s |
| Context window | BenchLeader | 1,000,000 tokens | 1,050,000 tokens |
| Reasoning | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 93.9% | 96.8% |
| SimpleBench | SimpleBench | 94.9%† | – |
| LiveBench 2.0 Reasoning | LiveBench | 88.8%† | 91.6% |
| Humanity's Last Exam (AA) | Artificial Analysis | 74.3% | – |
| HLE-Diamond (AA) | Artificial Analysis | 80.2% | 72.2% |
| ARC-AGI-1 | ARC Prize | 97.5% | 98.5% |
| ARC-AGI-2 | ARC Prize | 92.5% | 95.8% |
| ARC-AGI-3 | ARC Prize | 94.9% | 99.8% |
| CritPt | Artificial Analysis | 35.4% | – |
| CritPt v2 hard subset (AA) | Artificial Analysis | 76.5% | 71.5% |
| MysteryMechanism | Vals AI | 51.8%† | 53.1% |
| HLE-Diamond (Scale) | Scale AI SEAL | 72.4%† | 66.9%† |
| Chess Puzzles | Epoch AI Benchmarking Hub | 79.0% | 78.0% |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 90.0% | 95.0% |
| LMCA | Epoch AI Benchmarking Hub | 70.1%† | – |
| DTBench | Epoch AI Benchmarking Hub | 99.2%† | – |
| Coding | |||
| SciCode | SciCode | 67.7% | – |
| Arena WebDev | Arena (LMArena) | 1824† | 1781 |
| LiveBench 2.0 Coding | LiveBench | 79.2%† | 91.9% |
| Code Migration v1.1 | Vals AI | 83.3%† | – |
| ProgramBench | Vals AI | 19.0%† | – |
| SWE-bench Multimodal | Vals AI | 62.3%† | – |
| Vibe Code Bench v1.1 | Vals AI | 93.8%† | 89.9% |
| Vibe Code Bench 1-100 v1.1 | Vals AI | – | 55.2% |
| SWE Atlas v1.1 | Scale AI SEAL | 80.5%† | 60.5%† |
Show all 113 benchmarks
| Metric | Publisher | Claude Fable 5.2 (max) | GPT-6.2 Astra (max) |
|---|---|---|---|
| Coding | |||
| SWE Atlas v1.1: Codebase QnA | Scale AI SEAL | 83.9%† | 67.7%† |
| SWE Atlas v1.1: Test Writing | Scale AI SEAL | 77.8%† | 62.2%† |
| SWE Atlas v1.1: Refactoring | Scale AI SEAL | 80.0%† | 51.4%† |
| MirrorCode | Epoch AI Benchmarking Hub | – | 67.8% |
| CursorBench | Cursor | 60.0% | – |
| ALE-Bench | Epoch AI Benchmarking Hub | 2509.8† | – |
| FrontierSWE | Epoch AI Benchmarking Hub | 69.9% | – |
| WeirdML v3 | WeirdML | 43.1%† | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 59.6% | 64.1% |
| GameBench 2 Overall | DuelLab GameBench 2 | 95.0† | – |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | 99.2† | – |
| Design Arena Website | Design Arena | 1349† | 1339† |
| FrontierCode 1.1 (AA) | Artificial Analysis | 56.2% | – |
| FrontierCode 2.0 (AA) | Artificial Analysis | 48.8% | 45.4% |
| Agents & tools | |||
| Terminal-Bench | Terminal-Bench | 65.8% | 62.1% |
| Remote Labor Index | Scale AI / CAIS | 41.3%† | – |
| APEX-Agents | Mercor | 80.1%† | – |
| Arena Agent | Arena (LMArena) | 27.4 | 15.2 |
| LiveBench Long Agentic | LiveBench | 82.9%† | 70.8% |
| LiveBench Browser | LiveBench | 86.9%† | 93.0% |
| HiL-Bench | Scale AI SEAL | 80.0%† | – |
| GDPval-AA v2.1 | Artificial Analysis | 1831 | – |
| Analyst Agent (AA) | Artificial Analysis | – | 62.5% |
| Finance Agent v2.1 | Vals AI | 69.5%† | 60.8% |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | 5.0%† | 16.3% |
| Legal Research Bench | Vals AI | 46.1%† | 42.8% |
| Robo-bench | Vals AI | – | 32.7% |
| Tax Agent Bench | Vals AI | 32.1%† | 21.8% |
| Terminal-Bench 4.0 (Vals) | Vals AI | 65.2%† | 63.1% |
| Terminal-Bench Science | Vals AI | 55.7%† | 71.4% |
| Arena Agent Task Completion | Arena (LMArena) | 22.5 | 12.4 |
| PostTrainBench | Epoch AI Benchmarking Hub | 44.6% | – |
| GBAEval | Epoch AI Benchmarking Hub | 84.8%† | – |
| Vending-Bench 2 | Andon Labs | 9628.5† | – |
| SpireClimb | Epoch AI Benchmarking Hub | 15.3% | 41.3% |
| AutomationBench | Artificial Analysis | 76.4% | 77.3% |
| GDP.pdf (AA) | Artificial Analysis | 24.2% | 29.8% |
| Harvey LAB-AA v1.1 | Harvey | 7.2% | 15.0% |
| EnterpriseOps-Gym | Artificial Analysis | – | 29.9% |
| AA-Briefcase v1.1 | Artificial Analysis | 1785 | 1643 |
| GDPx | GDPx by SemiAnalysis | 16.4% | – |
| Design Arena Agentic | Design Arena | 1362† | 1304 |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | 73.3% | 78.1% |
| CWE-Bench (AA) | Artificial Analysis | – | 65.0% |
| DeepSec-Bench (AA) | Artificial Analysis | – | 46.8% |
| CyberGym end-to-end (AA) | Artificial Analysis | – | 0.0% |
| MKA (AA) | Artificial Analysis | 56.6% | 52.7% |
| Maths | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 95.1% | 92.6% |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 100.0% | 100.0% |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 99.7% | 100.0% |
| ProofBench | Vals AI | 100.0% | – |
| Autoform-Hard | Vals AI | 85.7%† | 87.5% |
| FrontierMath Erdős | Epoch AI Benchmarking Hub | – | 10.3% |
| Knowledge | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 94.0% | 78.5% |
| LiveBench 2.0 Data Analysis | LiveBench | 81.5%† | 84.1% |
| LiveBench Research | LiveBench | 94.5%† | 91.5% |
| AA-Omniscience v1.1 | Artificial Analysis | 66.8 | 50 |
| PRBench Finance | Scale AI SEAL | 47.1%† | – |
| PRBench Legal | Scale AI SEAL | 49.0%† | 43.1%† |
| BioMysteryBench | Vals AI | 83.7%† | 80.7% |
| Excel Modeling Benchmark v1.1 | Vals AI | 77.4%† | 76.0% |
| Immigration Bench | Vals AI | 74.0%† | – |
| DrugDiscoveryBench | Scale AI SEAL | – | 65.5%† |
| EBR-bench | Epoch AI Benchmarking Hub | 90.5% | 92.4% |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | 79.1% | 68.6% |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | 40.7% | 40.9% |
| HealthBench Professional (AA) | OpenAI | 67.7% | – |
| Instruction following | |||
| MultiChallenge | Scale AI SEAL | – | 73.2%† |
| Multimodal | |||
| LiveBench Design | LiveBench | 85.1%† | 82.8% |
| MMMU-Pro | Artificial Analysis | – | 89.5% |
| Humanity's Sixth Sense | Scale AI SEAL | 59.5% | 63.2%† |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 56.6%† | – |
| Furniture Assembly | Epoch AI Benchmarking Hub | 90.0% | 89.2% |
| VIALS-AA | Handshake AI | 67.1% | 65.8% |
| AA-3D | Artificial Analysis | 1271 | 1318 |
| Long context | |||
| AA-LCR v1.1 | Artificial Analysis | 83.3% | – |
| MLCR | Artificial Analysis | 43.9% | 39.4% |
| Composite | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 173.5† | 174.2† |
| LiveBench 2.0 | LiveBench | 84.9%† | 84.9% |
| AA Intelligence Index v5.0.1 | Artificial Analysis | 64.6 | 62.1 |
| AA Progress Index v1 | Artificial Analysis | 358 | – |
| Vals Index | Vals AI | 71.3† | 67.7 |
| Vals Finance Index | Vals AI | 63.6%† | 57.3% |
| Vals RSI Index | Vals AI | 37.9% | 29.0% |
| Vals Science Index | Vals AI | 63.7%† | 68.4% |
| Vals Legal Index | Vals AI | 69.2%† | 69.7% |
| AA Cyber Index | Artificial Analysis | – | 37.3 |
| Safety & honesty | |||
| FORTRESS | Scale AI SEAL | 15.6%† | – |
Data as of 2026-11-19. Best configuration of each model; every score links to its source on the model pages.
DuelLab GameBench 2, release gb2-20261114-ca113f475f5e (https://benchmarks.duellab.org/releases/gb2-20261114-ca113f475f5e/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.
What changed
- GPT-6.2 Astra (max): BenchLeader Index 67.8 → 68.9.
- Claude Fable 5.2 (max): 2 score revisions.
- GPT-6.2 Astra (max): 6 new results (VIALS-AA, Arena WebDev, Arena Agent…).
- GPT-6.2 Astra (max): BenchLeader Index 67.2 → 67.8.
- GPT-6.2 Astra (max): 2 new results (MKA (AA), Terminal-Bench).
- Claude Fable 5.2 (max): BenchLeader Index 72.7 → 71.8.
- Claude Fable 5.2 (max): 1 new result (PostTrainBench).
- Claude Fable 5.2 (max): 2 score revisions.
- Claude Fable 5.2 (max): 1 score revision.
- Claude Fable 5.2 (max): BenchLeader Index 74.9 → 72.7.
Related comparisons
- Claude Fable 5.2 vs Gemini 4 Carbon
- Claude Fable 5.2 vs Claude Opus 5.5
- Claude Fable 5.2 vs Gemini 4 Neon
- GPT-6.2 Astra vs Gemini 4 Carbon
- GPT-6.2 Astra vs Claude Opus 5.5
- GPT-6.2 Astra vs Gemini 4 Neon
- Claude Sonnet 5.5 vs GPT-6.2 Astra
- Claude Fable 5.2 vs GPT-6.1 Terra
Rankings: Best LLM overall · Best LLM for coding · Best LLM for agents · Best cheap LLMs
Claude Fable 5.2 vs GPT-6.2 Astra: questions
- Is Claude Fable 5.2 better than GPT-6.2 Astra?
- Claude Fable 5.2 (max) and GPT-6.2 Astra (max) are neck and neck on quality (71.5 vs 68.9, within the ±4.4 margin). The BenchLeader Index combines every independent quality benchmark; the two are within the margin of error as of 19 Nov 2026, so check the category scores for your use.
- Claude Fable 5.2 vs GPT-6.2 Astra: which is cheaper, and will that change?
- They cost the same: $16.00 per million tokens blended. No price change has been announced for either.
- Is Claude Fable 5.2 or GPT-6.2 Astra better for coding?
- Claude Fable 5.2 scores higher in coding (78 vs 76 on the category index, where 50 is average). On FrontierCode 2.0 (AA), the top shared coding benchmark, it is 48.8% for Claude Fable 5.2 against 45.4% for GPT-6.2 Astra.
- Which is faster, Claude Fable 5.2 or GPT-6.2 Astra?
- Claude Fable 5.2 streams ~2× faster (110 vs 49 output tokens per second, Artificial Analysis 7-day median).
- Is Claude Fable 5.2 or GPT-6.2 Astra better for agents and tool use?
- Claude Fable 5.2 scores higher in agentic tasks (75 vs 66 on the category index, where 50 is average). On Terminal-Bench, the top shared agents & tools benchmark, it is 65.8% for Claude Fable 5.2 against 62.1% for GPT-6.2 Astra.
More questions (4)
- Which is better for long documents?
- Both take about 1.1M tokens: Claude Fable 5.2 1,000,000, GPT-6.2 Astra 1,050,000.
- Which has lower latency?
- GPT-6.2 Astra answers first: 202.7 s against 221.3 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 Fable 5.2 vs GPT-6.2 Astra?
- Across the 33 benchmarks both have, Claude Fable 5.2 leads 17, GPT-6.2 Astra leads 15 and 1 is a tie. Every overall index that rates both puts Claude Fable 5.2 ahead.
- When was GPT-6.2 Astra released?
- GPT-6.2 Astra was released on 17 Nov 2026.