Anthropic: Claude Haiku Latest vs OpenAI: GPT-5.4
Side-by-side comparison of specs, pricing, benchmark scores, and task rankings. Updated 2026-11-07.
| Anthropic: Claude Haiku Latest | OpenAI: GPT-5.4 | |
|---|---|---|
| Vendor | ~anthropic | openai |
| Quality Score | 100 | 100 |
| Benchmark Score | - | 89.6 |
| Input Price | $0.10/M | $2.50/M |
| Output Price | $0.50/M | $15.00/M |
| Context Window | 1,000,000 | 1,050,000 |
| Max Output | 128,000 | 128,000 |
| Tool Calling | ✓ | ✓ |
| Structured Output | ✓ | ✓ |
| Reasoning Mode | ✓ | ✓ |
| Vision | ✓ | ✓ |
| Audio | - | - |
| Benchmark Scores | ||
| ai_index | - | 84.8 |
| ai_index_agentic | - | 67.8 |
| ai_index_coding | - | 100.0 |
| eqbench | - | 75.6 |
The verdict
For a sample job of 1 million input tokens and 250,000 output tokens, Claude Haiku Latest costs $0.23 and GPT-5.4 costs $6.25, so Claude Haiku Latest is about 27.8 times cheaper for the same work.
Both take in a similar amount of text: 1,000,000 and 1,050,000 tokens, about 1,500 and 1,575 pages.
Across the 20 tasks where they differ, Claude Haiku Latest scores higher on 3 (Code Completion, Bulk Data Labeling, Short-Form Summarization), GPT-5.4 on 17 (SQL Generation, Code Review, Code Refactoring and others).
Claude Haiku Latest was released in April 2026 and GPT-5.4 in March 2026.
Pick Claude Haiku Latest if you need lower cost; pick GPT-5.4 if you need the stronger all-round task scores.
Costs use current list prices per million tokens; page counts assume about 0.75 words per token and 500 words per page.
Who wins by task?
| Task | Anthropic: Claude Haiku Latest | OpenAI: GPT-5.4 |
|---|---|---|
| SQL Generation | 133 | 174 |
| Code Review | 132 | 175 |
| Code Completion | 131 | 120 |
| Code Refactoring | 136 | 173 |
| Bug Fixing | 136 | 187 |
| Unit Test Generation | 124 | 158 |
| Code Documentation | 131 | 146 |
| Regex Writing | 119 | 136 |
| CI/CD Pipelines | 120 | 149 |
| Frontend Component Design | 122 | 148 |
| Data Analysis | 124 | 173 |
| CSV / Spreadsheet Cleanup | 133 | 157 |
| ETL Scripting | 128 | 161 |
| JSON Extraction | 131 | 137 |
| Bulk Data Labeling | 129 | 122 |
| OCR / Document Parsing | 131 | 148 |
| Table Extraction from PDFs | 131 | 148 |
| Long-Document Summarization | 137 | 167 |
| Short-Form Summarization | 123 | 122 |
| Blog Post Writing | 121 | 143 |
Scores reflect capability match + benchmark data + pricing for each task. Methodology →