Compare / head-to-head
Claude Fable 5vsClaude Opus 4.6
Claude Fable 5 leads 26 of 28 shared benchmarks. Claude Opus 4.6 is 2x cheaper per token. Both offer a 1M-token context window.
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
26 – 2
Claude Fable 5 leads
Cheaper per token
Claude Opus 4.6
2x cheaper, input + output
Larger context
Tie
both 1M tokens
Providers
Anthropic
same provider
Claude Fable 5
Anthropic · released 2026-06-09
textvision
Claude Opus 4.6
Anthropic · released 2026-02-05
textvision
Quality
Benchmark matrix
28 shared · 16 only Claude Fable 5 · 10 only Claude Opus 4.6| Benchmark | Claude Fable 5 | Claude Opus 4.6 | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 28 | ||||
| SWE-bench Verified | 95% ↗ | 80.8% ↗ | +14.2 pt | Claude Fable 5 |
| GPQA Diamond | 85.9% ↗epoch run | 91.3% ↗ | -5.4 pt | Claude Opus 4.6 |
| LMArena Elo | 1504.3 ↗ | 1497.5 ↗ | +6.8 | Claude Fable 5 |
| APEX-Agents (Mercor) | 63.6% ↗ | 46.3% ↗ | +17.3 pt | Claude Fable 5 |
| Chess Puzzles (Epoch AI run) | 41% ↗epoch run | 17% ↗epoch run | +24 pt | Claude Fable 5 |
| EBR-bench (Epoch AI run) | 39.5% ↗epoch run | 12.7% ↗epoch run | +26.8 pt | Claude Fable 5 |
| FrontierMath Tier 4 v2 (Epoch AI run) | 90.2% ↗epoch run | 26.8% ↗epoch run | +63.4 pt | Claude Fable 5 |
| FrontierMath Tiers 1-3 v2 (Epoch AI run) | 87% ↗epoch run | 66% ↗epoch run | +21 pt | Claude Fable 5 |
| Furniture Assembly (Epoch AI run) | 35.8% ↗epoch run | 28.3% ↗epoch run | +7.5 pt | Claude Fable 5 |
| GSO Opt@1 (GSO) | 76.5% ↗ | 37.3% ↗ | +39.2 pt | Claude Fable 5 |
| LiveBench Agentic Coding (LiveBench) | 62.2% ↗ | 49% ↗ | +13.2 pt | Claude Fable 5 |
| LiveBench Coding (LiveBench) | 86% ↗ | 78.2% ↗ | +7.8 pt | Claude Fable 5 |
| LiveBench Data Analysis (LiveBench) | 80.5% ↗ | 69.9% ↗ | +10.6 pt | Claude Fable 5 |
| LiveBench Instruction Following (LiveBench) | 75.8% ↗ | 63.3% ↗ | +12.5 pt | Claude Fable 5 |
| LiveBench Language (LiveBench) | 90.7% ↗ | 83.3% ↗ | +7.4 pt | Claude Fable 5 |
| LiveBench Mathematics (LiveBench) | 96% ↗ | 89.3% ↗ | +6.7 pt | Claude Fable 5 |
| LiveBench Reasoning (LiveBench) | 89.7% ↗ | 88.7% ↗ | +1 pt | Claude Fable 5 |
| LMArena Vision (LMArena) | 1308.8 ↗ | 1299.4 ↗ | +9.4 | Claude Fable 5 |
| LMArena WebDev (LMArena) | 1625.3 ↗ | 1545.9 ↗ | +79.4 | Claude Fable 5 |
| Mystery Game Puzzles (Epoch AI run) | 52% ↗epoch run | 25% ↗epoch run | +27 pt | Claude Fable 5 |
| OSWorld-Verified | 85% ↗ | 72.7% ↗ | +12.3 pt | Claude Fable 5 |
| OTIS Mock AIME 2024-2025 (Epoch AI run) | 100% ↗epoch run | 94.4% ↗epoch run | +5.6 pt | Claude Fable 5 |
| SAGE (Vals AI) | 51.9% ↗ | 51.6% ↗ | +0.3 pt | Claude Fable 5 |
| SimpleBench (SimpleBench) | 81.9% ↗ | 67.6% ↗ | +14.3 pt | Claude Fable 5 |
| SimpleQA Verified | 70.7% ↗epoch run | 47% ↗epoch run | +23.7 pt | Claude Fable 5 |
| tau2-bench Banking Knowledge (Sierra) | 39.7% ↗ | 27.3% ↗ | +12.4 pt | Claude Fable 5 |
| Vending-Bench 2 (Andon Labs) | 5680.26 ↗ | 8017.59 ↗ | -2337.3 | Claude Opus 4.6 |
| WeirdML (Håvard Tveit Ihle) | 91.9% ↗ | 78% ↗ | +13.9 pt | Claude Fable 5 |
| Only Claude Fable 5 reports · 16 | ||||
| AutomationBench | 17.4% ↗ | not reported | — | — |
| Blueprint-Bench 2 | 38.6% ↗ | not reported | — | — |
| CursorBench | 72.9% ↗ | not reported | — | — |
| DeepSWE v1.1 (Datacurve) | 69.9% ↗ | not reported | — | — |
| FrontierCode (Diamond) | 29.3% ↗ | not reported | — | — |
| FrontierSWE V2 (Proximal Labs) | 47% ↗ | not reported | — | — |
| GDP.pdf | 29.8% ↗ | not reported | — | — |
| Legal Agent Benchmark (Harvey's Held-Out Set) | 13.3% ↗ | not reported | — | — |
| LMArena Agent (LMArena) | 0.0821 ↗ | not reported | — | — |
| MirrorCode (Epoch AI run) | 63.9% ↗epoch run | not reported | — | — |
| OfficeQA Pro | 57.9% ↗ | not reported | — | — |
| OSWorld-Verified (XLANG) | 86% ↗ | not reported | — | — |
| SWE-Bench Pro | 80% ↗ | not reported | — | — |
| SWE-rebench 2026-05-15 to 2026-07-01 (Nebius) | 64.5% ↗ | not reported | — | — |
| Terminal-Bench 2.1 | 84.3% ↗ | not reported | — | — |
| Terminal-Bench 4.0 (Vals AI) | 41.4% ↗ | not reported | — | — |
| Only Claude Opus 4.6 reports · 10 | ||||
| AIME 2026 | not reported | 96.7% ↗matharena | — | — |
| ARC-AGI-2 (Verified) | not reported | 68.8% ↗ | — | — |
| MCP Atlas | not reported | 59.5% ↗ | — | — |
| MMMLU | not reported | 91.1% ↗ | — | — |
| MMMU-Pro (no tools) | not reported | 73.9% ↗ | — | — |
| MMMU-Pro (with tools) | not reported | 77.3% ↗ | — | — |
| SWE-bench Verified (Epoch AI run) | not reported | 78.7% ↗epoch run | — | — |
| Tau2-bench Retail | not reported | 91.9% ↗ | — | — |
| Tau2-bench Telecom | not reported | 99.3% ↗ | — | — |
| Terminal-Bench 2.0 | not reported | 65.4% ↗ | — | — |
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is Claude Fable 5 minus Claude Opus 4.6 in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing
Side by side
Official pricing pages and model cards| Spec | Claude Fable 5 | Claude Opus 4.6 | Edge |
|---|---|---|---|
| Context window | 1M tokens | 1M tokens | Tie |
| Max output | 128K tokens | 128K tokens | Tie |
| Input price / 1M | $10 ↗ | $5 ↗ | Claude Opus 4.6 |
| Output price / 1M | $50 ↗ | $25 ↗ | Claude Opus 4.6 |
| Cached input / 1M | $1 ↗ | $0.5 ↗ | Claude Opus 4.6 |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | $60 | $30 | Claude Opus 4.6 |
| Modalities | text · vision | text · vision | Tie |
| Released | 2026-06-09 | 2026-02-05 | — |
| Cited benchmark scores | 44 | 46 | — |
Reliability
Anthropic status
Live from /statusMore matchups
Claude Fable 5 vs …
Models sharing the most benchmarksMore matchups
Claude Opus 4.6 vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run Claude Fable 5 and Claude Opus 4.6 on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.