BlogLLM Benchmark Scores as Published, Side by Side
AlpacaEval: Results as Published
Every result the catalog holds for AlpacaEval: 5 models from 3 developers, read from 1 page. Each number is as printed by the company named beside it; nothing is ranked or rescored. Models are listed by developer, newest first.
Area Writing and ConversationMaintainer Stanford (Yann Dubois and co-authors)Introduced 2023Data version 2026103001, as of 2026-10-30
AlpacaEval scores an instruction-following model by letting an LLM judge compare its answers with a baseline model's answers on a fixed prompt set. The length-controlled win rate (AlpacaEval 2.0 LC) removes the judge's preference for longer answers by predicting what the win rate would be at equal length.
AlpacaEval 2.0 (length-controlled win rate)
| Model | Value | Printed by | Setup as printed |
|---|---|---|---|
| o1-miniOpenAI, released 2024-09-12 | 57.8% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
| GPT-4oOpenAI, released 2024-05-13 | 51.1% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
| Claude 3.5 SonnetAnthropic, released 2024-10-22 | 52% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
| DeepSeek-R1DeepSeek, released 2025-01-20 | 87.6% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
| DeepSeek-V3DeepSeek, released 2024-12-26 | 70% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.