Claude Haiku 5.5 vs Claude Sonnet 5.5
| Spec | Claude Haiku 5.5 | Claude Sonnet 5.5 |
|---|---|---|
| Respan Index | 57.2 (#16) | 73.1 (#2) |
| Blended price per 1M | $0.1 | $2 |
| Context window | 1M | 1M |
| Input / output per 1M | $0.05 / $0.25 | $1 / $5 |
| Max output | 128K | 128K |
| Input | text, vision | text, vision |
| Released | 2026-10-07 | 2026-09-28 |
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 Haiku 5.5 | Claude Sonnet 5.5 |
|---|---|---|
| Coding | 55.0 Claude Haiku 5.5 | 79.6 Claude Sonnet 5.5 |
| Agents | 65.9 Claude Haiku 5.5 | 78.4 Claude Sonnet 5.5 |
| Reasoning | 65.0 Claude Haiku 5.5 | 79.0 Claude Sonnet 5.5 |
| Knowledge & language | 23.6 Claude Haiku 5.5 | 48.5 Claude Sonnet 5.5 |
| Human preference | 55.5 Claude Haiku 5.5 | 60.2 Claude Sonnet 5.5 |
Benchmarks both report
Claude Sonnet 5.5 leads on 28 of the 31 benchmarks both report. Higher is better. Each score links to where it was published.
| Benchmark | Claude Haiku 5.5 | Claude Sonnet 5.5 |
|---|---|---|
| MirrorCode (Epoch AI) | 78.9% | 97.8% |
| SAGE 1.1 (Vals AI) | 89.8% | 88% |
| APEX-Agents (Mercor) | 56.1% | 75.5% |
| Chess Puzzles (Epoch AI) | 55% | 59% |
| LMArena Agent (LMArena) | 0.0143 | 0.1384 |
| LMArena Elo | 1425 | 1477 |
| LMArena WebDev (LMArena) | 1589 | 1770 |
| Terminal-Bench 4.0 (Vals AI) | 36.4% | 64.1% |
| Vending-Bench 2 (Andon Labs) | 3602.46 | 11165.28 |
| FrontierMath Tier 4 v2 (Epoch AI) | 46.3% | 80.5% |
| FrontierMath Tiers 1-3 v2 (Epoch AI) | 75.1% | 88.8% |
| FrontierSWE V2 (Proximal Labs) | 43.8% | 61.9% |
| Furniture Assembly (Epoch AI) | 47.5% | 75% |
| GPQA Diamond | 89.6% | 95.6% |
| LiveBench Agentic Coding (LiveBench) | 57.2% | 56.3% |
| LiveBench Coding (LiveBench) | 77.9% | 91.4% |
| LiveBench Data Analysis (LiveBench) | 72.8% | 78.6% |
| LiveBench Instruction Following (LiveBench) | 66.5% | 70.5% |
| LiveBench Language (LiveBench) | 63.5% | 83.4% |
| LiveBench Mathematics (LiveBench) | 92.8% | 96.7% |
| LiveBench Reasoning (LiveBench) | 84.6% | 91.6% |
| Mystery Game Puzzles (Epoch AI) | 30% | 65% |
| OTIS Mock AIME 2024-2025 (Epoch AI) | 98.9% | 100% |
| SAGE (Vals AI) | 45.8% | 51.8% |
| SimpleQA Verified | 23.8% | 46.5% |
| Chartography (no tools) | 46.4% | 61.6% |
| FrontierCode v1.1 (Main) | 46.4% | 46.2% |
| Humanity's Last Exam (no tools) | 45.9% | 56.9% |
| Humanity's Last Exam (with tools) | 57.4% | 64.5% |
| OSWorld 2.1 (offline subset) | 72.4% | 83.9% |
| Terminal-Bench 4.0 | 39.2% | 70.6% |
FAQ
Is Claude Haiku 5.5 better than Claude Sonnet 5.5?
Claude Sonnet 5.5 scores 73.1 on the Respan Index (#2), ahead of Claude Haiku 5.5 at 57.2 (#16). Claude Sonnet 5.5 leads on 28 of the 31 benchmarks both report.
Which is cheaper, Claude Haiku 5.5 or Claude Sonnet 5.5?
Claude Haiku 5.5 lists at $0.05 input and $0.25 output per 1M tokens; Claude Sonnet 5.5 at $1 and $5. Claude Haiku 5.5 is 20x cheaper per token at list price.
Which has the larger context window?
Both accept 1M tokens of context.