Claude Sonnet 4.6 vs Claude Sonnet 5
| Spec | Claude Sonnet 4.6 | Claude Sonnet 5 |
|---|---|---|
| Respan Index | 37.3 (#73) | 48.1 (#47) |
| Blended price per 1M | $6 | $4 |
| Context window | 1M | 1M |
| Input / output per 1M | $3 / $15 | $2 / $10 |
| Max output | 128K | 128K |
| Input | text, vision | text, vision |
| Released | 2026-02-17 | 2026-06-30 |
Respan Index by category
Each category is scored 0 to 100, and how far ahead the winner is. Leads under a point are close to even.
| Category | Claude Sonnet 4.6 | Claude Sonnet 5 |
|---|---|---|
| Coding | 14.9 Claude Sonnet 4.6 | 32.9 Claude Sonnet 5 |
| Agents | 50.9 Claude Sonnet 4.6 | 62.2 Claude Sonnet 5 |
| Reasoning | 36.5 Claude Sonnet 4.6 | 50.6 Claude Sonnet 5 |
| Knowledge & language | 35.7 Claude Sonnet 4.6 | 34.1 Claude Sonnet 5 |
| Human preference | 54.5 Claude Sonnet 4.6 | 54.9 Claude Sonnet 5 |
Benchmarks both report
Claude Sonnet 5 leads on 17 of the 23 benchmarks both report. Higher is better. Each score links to where it was published.
| Benchmark | Claude Sonnet 4.6 | Claude Sonnet 5 |
|---|---|---|
| LMArena Elo | 1472 | 1461 |
| LMArena WebDev (LMArena) | 1522 | 1539 |
| LMArena Vision (LMArena) | 1275 | 1261 |
| Chess Puzzles (Epoch AI) | 23% | 45% |
| APEX-Agents (Mercor) | 43% | 54.5% |
| DeepSWE v1.1 (Datacurve) | 29.9% | 53.8% |
| GPQA Diamond | 89.9% | 90.5% |
| LiveBench Agentic Coding (LiveBench) | 42.6% | 59.4% |
| LiveBench Coding (LiveBench) | 79.3% | 80.7% |
| LiveBench Data Analysis (LiveBench) | 77.9% | 71.7% |
| LiveBench Instruction Following (LiveBench) | 63.2% | 63.9% |
| LiveBench Language (LiveBench) | 76.1% | 75% |
| LiveBench Mathematics (LiveBench) | 87% | 92.9% |
| LiveBench Reasoning (LiveBench) | 84.8% | 88.7% |
| Mystery Game Puzzles (Epoch AI) | 16% | 35% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 85.8% | 94.7% |
| SAGE (Vals AI) | 46.6% | 48.9% |
| SimpleQA Verified | 35.5% | 33.7% |
| Vending-Bench 2 (Andon Labs) | 7204.14 | 6377.7 |
| WeirdML (Håvard Tveit Ihle) | 66.1% | 69% |
| Humanity's Last Exam (no tools) | 33.2% | 43.2% |
| Humanity's Last Exam (with tools) | 49% | 57.4% |
| SWE-bench Verified | 79.6% | 85.2% |
FAQ
Is Claude Sonnet 4.6 better than Claude Sonnet 5?
Claude Sonnet 5 scores 48.1 on the Respan Index (#47), ahead of Claude Sonnet 4.6 at 37.3 (#73). Claude Sonnet 5 leads on 17 of the 23 benchmarks both report.
Which is cheaper, Claude Sonnet 4.6 or Claude Sonnet 5?
Claude Sonnet 4.6 lists at $3 input and $15 output per 1M tokens; Claude Sonnet 5 at $2 and $10. Claude Sonnet 5 is 1.5x cheaper per token at list price.
Which has the larger context window?
Both accept 1M tokens of context.
Compare other models
- vsClaude Sonnet 4.6 vs Claude Sonnet 4.5
- vsClaude Sonnet 5.5 vs Claude Sonnet 5
- vsGemini 3.8 Flash vs Claude Sonnet 5
- vsGemini 4 Argon vs Claude Sonnet 5
- vsClaude Sonnet 4.6 vs Beam 1.0
- vsClaude Sonnet 4.6 vs GLM-5.1
- vsClaude Sonnet 4.6 vs GPT-5
- vsClaude Sonnet 5 vs GPT-5.6 Luna
- vsClaude Sonnet 5 vs Hy4 preview
- vsClaude Sonnet 5 vs GPT-6 Luna