Choosing a Coding Agent Model: Claude Opus 5, GPT-5.6 Sol, or Gemini 3.7 Flash

Description and use

<p>There is no benchmark result that settles this choice for every coding agent.</p> <p>The published <strong>Terminal-Bench 2.1</strong> numbers are close: <strong>Claude Opus 5 scores 89% at Max effort</strong>, <strong>GPT-5.6 Sol scores 88.8% in the reported single-agent run and 91.9% in Ultra</strong>, and <strong>Gemini 3.7 Flash scores 85.8%</strong>. Those results are useful for narrowing the field, but they are not a clean ranking. The reasoning effort, harness configuration, and Ultra's multi-agent setup differ.</p> <p>For a fixed request containing <strong>20,000 input tokens and 2,

Problem

<p>There is no benchmark result that settles this choice for every coding agent.</p> <p>The published <strong>Terminal-Bench 2.1</strong> numbers are close: <strong>Claude Opus 5 scores 89% at Max effort</strong>, <strong>GPT-5.6 Sol scores 88.8% in the reported single-agent run and 91.9% in Ultra</strong>, and <strong>Gemini 3.7 Flash scores 85.8%</strong>. Those results are useful for narrowing the field, but they are not a clean ranking. The reasoning effort, harness configuration, and Ultra's multi-agent setup differ.</p> <p>For a fixed request containing <strong>20,000 input tokens and 2,

Outcome

<p>There is no benchmark result that settles this choice for every coding agent.</p> <p>The published <strong>Terminal-Bench 2.1</strong> numbers are close: <strong>Claude Opus 5 scores 89% at Max effort</strong>, <strong>GPT-5.6 Sol scores 88.8% in the reported single-agent run and 91.9% in Ultra</strong>, and <strong>Gemini 3.7 Flash scores 85.8%</strong>. Those results are useful for narrowing the field, but they are not a clean ranking. The reasoning effort, harness configuration, and Ultra's multi-agent setup differ.</p> <p>For a fixed request containing <strong>20,000 input tokens and 2,

Evidence

Choosing a Coding Agent Model: Claude Opus 5, GPT-5.6 Sol, or Gemini 3.7 Flash

Source