Benchmarking Agents for Sapient
Compare coding agents across General, MCP, and CLI interfaces using the latest public custom-eval results and configured workflows.
Agent performance
112 of 163 runs pass / 69%Updated Sep 23, 2026
| Agent | Total | Response Validation | Sapient CLI Custom Prompts | Discovery Test | Test | documentation-quality | API performance | Pricing Clarity | Other |
|---|---|---|---|---|---|---|---|---|---|
1  Anthropic / Claude Sonnet 4.6 | |||||||||
2 Claude Code / Claude Opus 4.6 | — | — | — | — | — | — | — | ||
3 Claude Code / Claude Sonnet 4.6 | — | — | — | — | — | ||||
4 Claude Code / Claude Haiku 4.5 | — | — | — | ||||||
5  Codex CLI / GPT-5.4 Mini | — | — | — | — | — | — | — | ||
6 Codex CLI / GPT-5.5 | — | — | — | — | — | — | |||
7 Codex CLI / GPT-5.4 | — | — | — | — | — | ||||
8  Cursor / Auto | — | — | — | — | — | — | — | ||
9  OpenAI / GPT-5 Nano | — | — | — | — | — | — | — |