PhysDesignBench
GitHub
Design task · v1

MZI-v1

Broadband 4x4 Rectangular MZI Mesh

Design a broadband 4×4 rectangular MZI mesh with independently controlled interferometers and trusted optical, electrical, and thermal verification.

Runs
13
Scored finals
8 of 13
Best final
S = 0.72 GPT-6.1 Sol · Codex
Score over timeBest-so-far score S for each run · small dots are probes, large dots final submissions
Score S from 0 to 1 against elapsed minutes. Final scores: GPT-6.1 Sol (Codex) 0.722; Sonnet 5.5 (high) (Claude Code) 0.681; Opus 5.5 (Claude Code) 0.671; Sonnet 5.5 (mini-SWE-agent) 0.638; GPT-6 Astra (Codex) 0.541; GPT-5.6 Sol (Codex) 0.326; Fable 5.1 (Claude Code) 0.321; GPT-6.1 Sol (mini-SWE-agent) 0.294; Gemini 4 Argon (mini-SWE-agent) Invalid; GLM 5.3 Flash (high) (Claude Code) Invalid; GLM 5.3 Flash (max) (mini-SWE-agent) Invalid; Kimi K3.1 Pro (Claude Code) Invalid; Muse Wonder (high) (Muse Code) Invalid.0.000.250.500.751.000153045607590105120135150Elapsed time (min)Score SGPT-6.1 Sol · mini-SWE-agent: final S = 0.294 at 125 minSonnet 5.5 · mini-SWE-agent: final S = 0.638 at 137 minGemini 4 Argon · mini-SWE-agent: final Invalid at 121 minGLM 5.3 Flash (high) · Claude Code: final Invalid at 102 minGLM 5.3 Flash (max) · mini-SWE-agent: final Invalid at 86 minKimi K3.1 Pro · Claude Code: final Invalid at 121 minMuse Wonder (high) · Muse Code: final Invalid at 121 minFable 5.1 · Claude Code: final S = 0.321 at 124 minGPT-5.6 Sol · Codex: final S = 0.326 at 124 minGPT-6 Astra · Codex: final S = 0.541 at 143 minOpus 5.5 · Claude Code: final S = 0.671 at 120 minSonnet 5.5 (high) · Claude Code: final S = 0.681 at 112 minGPT-6.1 Sol · Codex: final S = 0.722 at 147 min

Showing 13 of 13 runs. Hover a line or a model to isolate it; click a model to hide or show it.

Final layouts

Every run’s final GDS, at one scale.

The design each run ended with, drawn from its evaluated GDS. Every layout uses the same physical scale, so sizes compare directly. Hatched layouts failed a validity gate and score zero. Select a layout to zoom and toggle its layers.

  • Optical mask
  • Heater
  • Metal
  • Pad opening
  • Phase marker
  • MMI marker

  • MZI-v1 · Codex · GPT-6.1-sol

    S = 0.722 · scale bar 200 µm

    Fit
    GPT-6.1 SolCodexS = 0.72

  • MZI-v1 · Claude Code · Claude Sonnet 5.5

    S = 0.681 · scale bar 200 µm

    Fit
    Sonnet 5.5 (high)Claude CodeS = 0.68

  • MZI-v1 · Claude Code · Claude Opus 5.5

    S = 0.671 · scale bar 200 µm

    Fit
    Opus 5.5Claude CodeS = 0.67

  • MZI-v1 · mini-SWE-agent · Claude Sonnet 5.5

    S = 0.638 · scale bar 200 µm

    Fit
    Sonnet 5.5mini-SWE-agentS = 0.64

  • MZI-v1 · Codex · GPT-6 Astra

    S = 0.541 · scale bar 200 µm

    Fit
    GPT-6 AstraCodexS = 0.54

  • MZI-v1 · Codex · GPT-5.6-sol

    S = 0.326 · scale bar 200 µm

    Fit
    GPT-5.6 SolCodexS = 0.33

  • MZI-v1 · Claude Code · Claude Fable 5.1

    S = 0.321 · scale bar 200 µm

    Fit
    Fable 5.1Claude CodeS = 0.32

  • MZI-v1 · mini-SWE-agent · GPT-6.1-sol

    S = 0.294 · scale bar 200 µm

    Fit
    GPT-6.1 Solmini-SWE-agentS = 0.29
  • No final GDS: the evaluator rejected the final source before producing a layout
    Gemini 4 Argonmini-SWE-agentInvalid · S = 0
  • No final GDS: the evaluator rejected the final source before producing a layout
    GLM 5.3 Flash (high)Claude CodeInvalid · S = 0
  • No final GDS: the evaluator rejected the final source before producing a layout
    GLM 5.3 Flash (max)mini-SWE-agentInvalid · S = 0

  • MZI-v1 · Claude Code · Kimi K3.1 Pro

    Invalid final · S = 0 · scale bar 200 µm

    Fit
    Kimi K3.1 ProClaude CodeInvalid · S = 0

  • MZI-v1 · Muse Code · Muse Wonder

    Invalid final · S = 0 · scale bar 200 µm

    Fit
    Muse Wonder (high)Muse CodeInvalid · S = 0