Coding
Writing and editing code against a specification, measured without an agent loop. Distinct from agentic coding: these benchmarks score the model's output directly rather than a model-plus-harness system.
Ordered by how many of these benchmarks place a model in their top 10, then by its best placing. Each column keeps its own order; nothing here is averaged.
| Model | In top 10 | Artificial Analysis LLM Leaderboard SciCode | Artificial Analysis LLM Leaderboard Coding Index | Arena — Code Overall |
|---|---|---|---|---|
 Claude Fable 5 (Opus 4.8 fallback) anthropic | 3/3 | #1 60.2% | #5 76.5 | #4 1626 |
 GPT-5.6 Sol openai | 3/3 | #5 56.9% | #1 78.3 | #5 1619 |
| Claude Opus 5 anthropic | 3/3 | #8 55.7% | #2 78.0 | #1 1691 |
 Kimi K3 kimi | 3/3 | #3 58.7% | #6 76.2 | #2 1674 |
 Grok 4.6 xai | 3/3 | #9 54.6% | #3 76.8 | #3 1629 |
 Gemini 3.7 Flash google | 3/3 | #4 57.9% | #7 76.1 | #7 1587 |
 GLM-5.3 zai | 3/3 | #6 56.5% | #8 74.8 | #6 1599 |
 Muse Spark 1.2 meta | 2/3 | #7 56.4% | #10 72.2 | #16 1534 |
| Grok 4.5 xai | 2/3 | #10 54.1% | #9 72.4 | #11 1556 |
| Gemini 3.1 Pro Preview google | 1/3 | #2 58.9% | #18 68.8 | #31 1446 |
| GPT-5.6 Terra openai | 1/3 | #11 53.9% | #4 76.7 | #18 1520 |
| GLM-5.2 zai | 1/3 | #19 50.5% | #19 68.8 | #8 1582 |
| Claude Opus 4.8 anthropic | 1/3 | — | — | #9 1563 |
| Claude Opus 4.7 anthropic | 1/3 | — | — | #10 1558 |
A dash means the source does not cover that model, not that it scored zero. Models with no registry entry are excluded here because they cannot be joined across sources — they are still shown, flagged, on their own source's board.