---
title: "claude-code \u00b7 Sonnet 4.6 | Pelican benchmark"
url: https://evaljetty.com/pelicans/runs/agents/claude-sonnet.html
description: "claude-code \u00b7 Sonnet 4.6 (claude-sonnet-4-6) drawing a pelican riding a bicycle: every hill-climb run with self-scores, external judge scores and Jetty trajectories."
updated: 2026-09-28
publisher: Jetty (https://jetty.io)
---

02 — Runs · v1 · jettyio · hand-curated runbook edits

# claude-code claude-sonnet-4-6

The reference lineage: Jon edited the runbook by hand eleven times, reading each trajectory before writing the next version.

Judge, best round

26.2/40

round v5 (agent's pick)

Self-score

37/40

inflation +10.8

Runs

11

not recorded (v1)

![claude-code · Sonnet 4.6 best round](https://evaljetty.com/pelicans/img/claude-sonnet/v5.svg)

best round v5 · 7ddc3c19

### Self-score vs judge, by axis

Best round v5, 0–10 per axis

agent self-scoreexternal judge

**Judge:** The pelican reads as a pelican thanks to the large orange pouch and white body. The beak is a boxy block, the head has odd buck-teeth-like shapes, and the wings are messy. The bicycle has spoked wheels, a plausible frame, a handlebar and a crank, but the seat is hidden and the fork and frame geometry are slightly off. The pelican sits on the frame with long legs reaching the pedals and a wing gripping the handlebar, which sells the riding pose. Clean colors and background, though the head and pouch shapes are crude.

### Score by round

Out of 40. Each round starts from the previous round's best SVG. Dashed line: the 28/40 pass threshold.

self-scorejudgepass ≥ 28

## 11 drawings, *oldest first.*

Click a drawing to open it in the head-to-head viewer. The outlined card is the agent's best round. Verdicts are derived (code checks on the stored SVG and judge ≥ 28/40); see [Runs](https://evaljetty.com/pelicans/runs/index.html#all-runs).

[![claude-code · Sonnet 4.6 round v1](https://evaljetty.com/pelicans/img/claude-sonnet/v1.svg)](https://evaljetty.com/pelicans/investigations/best-pelican.html#claude-sonnet/v1)

**v1** failjudge **26.5** · self 31
7ddfcf08[runbook](https://evaljetty.com/pelicans/author/versions.html#claude-sonnet/v1)

[![claude-code · Sonnet 4.6 round v2](https://evaljetty.com/pelicans/img/claude-sonnet/v2.svg)](https://evaljetty.com/pelicans/investigations/best-pelican.html#claude-sonnet/v2)

**v2** failjudge **25.2** · self 32
414e644d[runbook](https://evaljetty.com/pelicans/author/versions.html#claude-sonnet/v2)

[![claude-code · Sonnet 4.6 round v3](https://evaljetty.com/pelicans/img/claude-sonnet/v3.svg)](https://evaljetty.com/pelicans/investigations/best-pelican.html#claude-sonnet/v3)

**v3** failjudge **24.5** · self 34
b169371d[runbook](https://evaljetty.com/pelicans/author/versions.html#claude-sonnet/v3)

[![claude-code · Sonnet 4.6 round v4](https://evaljetty.com/pelicans/img/claude-sonnet/v4.svg)](https://evaljetty.com/pelicans/investigations/best-pelican.html#claude-sonnet/v4)

**v4** failjudge **24.5** · self 36
063dfed9[runbook](https://evaljetty.com/pelicans/author/versions.html#claude-sonnet/v4)

[![claude-code · Sonnet 4.6 round v5](https://evaljetty.com/pelicans/img/claude-sonnet/v5.svg)](https://evaljetty.com/pelicans/investigations/best-pelican.html#claude-sonnet/v5)

**v5** · best failjudge **26.2** · self 37
7ddc3c19[runbook](https://evaljetty.com/pelicans/author/versions.html#claude-sonnet/v5)

[![claude-code · Sonnet 4.6 round v6](https://evaljetty.com/pelicans/img/claude-sonnet/v6.svg)](https://evaljetty.com/pelicans/investigations/best-pelican.html#claude-sonnet/v6)

**v6** failjudge **26.5** · self 36
39febb11[runbook](https://evaljetty.com/pelicans/author/versions.html#claude-sonnet/v6)

[![claude-code · Sonnet 4.6 round v7](https://evaljetty.com/pelicans/img/claude-sonnet/v7.svg)](https://evaljetty.com/pelicans/investigations/best-pelican.html#claude-sonnet/v7)

**v7** failjudge **27.7** · self 34
ddf779bd[runbook](https://evaljetty.com/pelicans/author/versions.html#claude-sonnet/v7)

[![claude-code · Sonnet 4.6 round v8](https://evaljetty.com/pelicans/img/claude-sonnet/v8.svg)](https://evaljetty.com/pelicans/investigations/best-pelican.html#claude-sonnet/v8)

**v8** failjudge **26.3** · self 34
79fdee62[runbook](https://evaljetty.com/pelicans/author/versions.html#claude-sonnet/v8)

[![claude-code · Sonnet 4.6 round v9](https://evaljetty.com/pelicans/img/claude-sonnet/v9.svg)](https://evaljetty.com/pelicans/investigations/best-pelican.html#claude-sonnet/v9)

**v9** failjudge **25.8** · self 33
5f980543[runbook](https://evaljetty.com/pelicans/author/versions.html#claude-sonnet/v9)

[![claude-code · Sonnet 4.6 round v10](https://evaljetty.com/pelicans/img/claude-sonnet/v10.svg)](https://evaljetty.com/pelicans/investigations/best-pelican.html#claude-sonnet/v10)

**v10** passjudge **28** · self 35
2043d7e3[runbook](https://evaljetty.com/pelicans/author/versions.html#claude-sonnet/v10)

[![claude-code · Sonnet 4.6 round v11](https://evaljetty.com/pelicans/img/claude-sonnet/v11.svg)](https://evaljetty.com/pelicans/investigations/best-pelican.html#claude-sonnet/v11)

**v11** failjudge **25.3** · self 35
3a4c3357[runbook](https://evaljetty.com/pelicans/author/versions.html#claude-sonnet/v11)

### Scores per round

| Round | Pelican | Bicycle | Composition | Polish | Self total | Judge total | Verdict | Judge spread | Cost | Minutes | Trajectory |
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |
| v1 | 6.7self 8 | 6.5self 8 | 6.7self 7 | 6.7self 8 | 31 | 26.5 | fail | 1 | – | 7 | 7ddfcf08 |
| v2 | 5.7self 8 | 6.7self 8 | 6.7self 8 | 6.2self 8 | 32 | 25.2 | fail | 1 | – | 12.6 | 414e644d |
| v3 | 5.7self 9 | 6.2self 9 | 6.7self 8 | 6self 8 | 34 | 24.5 | fail | 1.5 | – | 10.3 | b169371d |
| v4 | 5.3self 9 | 6.7self 9 | 6.7self 9 | 5.8self 9 | 36 | 24.5 | fail | 2.5 | – | 7 | 063dfed9 |
| v5 | 5.8self 9 | 7self 9 | 7self 10 | 6.3self 9 | 37 | 26.2 | fail | 1 | – | 7.8 | 7ddc3c19 |
| v6 | 6.2self 9 | 6.8self 9 | 7self 9 | 6.5self 9 | 36 | 26.5 | fail | 3 | – | 5.4 | 39febb11 |
| v7 | 6.8self 8 | 6.8self 9 | 7.5self 9 | 6.5self 8 | 34 | 27.7 | fail | 0.5 | – | 13.8 | ddf779bd |
| v8 | 6self 8 | 6.8self 9 | 7self 8 | 6.5self 9 | 34 | 26.3 | fail | 3.5 | – | 5.3 | 79fdee62 |
| v9 | 6self 8 | 6.7self 9 | 6.7self 7 | 6.5self 9 | 33 | 25.8 | fail | 1 | – | 7.3 | 5f980543 |
| v10 | 6.8self 9 | 7.2self 9 | 7.2self 8 | 6.8self 9 | 35 | 28 | pass | 1.5 | – | 6.7 | 2043d7e3 |
| v11 | 6self 9 | 6.7self 9 | 6.3self 8 | 6.3self 9 | 35 | 25.3 | fail | 1 | – | 9.5 | 3a4c3357 |

---
Source: https://evaljetty.com/pelicans/runs/agents/claude-sonnet.html · Evals by Jetty · Run your own evals: https://jetty.io/?utm_source=evaljetty&utm_medium=llms
