PhysDesignBench
GitHub
Design task · v1

Serial MZI-v1

Serial Dual-Ring-Assisted MZI Elliptic Bandpass Electrothermal Design v1

Design a serial dual-ring-assisted MZI elliptic bandpass filter with five tunable couplers, three electrothermal phase controls, and trusted optical, electrical, thermal, DRC, and circuit verification.

Runs
13
Scored finals
8 of 13
Best final
S = 0.30 Gemini 4 Argon · mini-SWE-agent
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: Gemini 4 Argon (mini-SWE-agent) 0.304; GPT-6.1 Sol (Codex) 0.102; Kimi K3.1 Pro (Claude Code) 0.099; Sonnet 5.5 (mini-SWE-agent) 0.062; GPT-6 Astra (Codex) 0.047; Fable 5.1 (Claude Code) 0.041; Opus 5.5 (Claude Code) 0.039; Muse Wonder (high) (Muse Code) 0.038; GLM 5.3 Flash (high) (Claude Code) Invalid; GLM 5.3 Flash (max) (mini-SWE-agent) Invalid; GPT-5.6 Sol (Codex) Invalid; GPT-6.1 Sol (mini-SWE-agent) Invalid; Sonnet 5.5 (high) (Claude Code) Invalid.0.000.250.500.751.000306090120150180Elapsed time (min)Score SGPT-6.1 Sol · mini-SWE-agent: final Invalid at 142 minSonnet 5.5 · mini-SWE-agent: final S = 0.062 at 122 minGLM 5.3 Flash (high) · Claude Code: final Invalid at 116 minGLM 5.3 Flash (max) · mini-SWE-agent: final Invalid at 141 minGPT-5.6 Sol · Codex: final Invalid at 161 minSonnet 5.5 (high) · Claude Code: final Invalid at 124 minMuse Wonder (high) · Muse Code: final S = 0.038 at 117 minOpus 5.5 · Claude Code: final S = 0.039 at 66 minFable 5.1 · Claude Code: final S = 0.041 at 146 minGPT-6 Astra · Codex: final S = 0.047 at 121 minKimi K3.1 Pro · Claude Code: final S = 0.099 at 121 minGPT-6.1 Sol · Codex: final S = 0.102 at 121 minGemini 4 Argon · mini-SWE-agent: final S = 0.304 at 143 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

  • Serial MZI-v1 · mini-SWE-agent · Gemini-4-Argon

    S = 0.304 · scale bar 100 µm

    Fit
    Gemini 4 Argonmini-SWE-agentS = 0.30

  • Serial MZI-v1 · Codex · GPT-6.1-sol

    S = 0.102 · scale bar 100 µm

    Fit
    GPT-6.1 SolCodexS = 0.10

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

    S = 0.099 · scale bar 100 µm

    Fit
    Kimi K3.1 ProClaude CodeS = 0.10

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

    S = 0.062 · scale bar 100 µm

    Fit
    Sonnet 5.5mini-SWE-agentS = 0.06

  • Serial MZI-v1 · Codex · GPT-6 Astra

    S = 0.047 · scale bar 100 µm

    Fit
    GPT-6 AstraCodexS = 0.05

  • Serial MZI-v1 · Claude Code · Claude Fable 5.1

    S = 0.041 · scale bar 100 µm

    Fit
    Fable 5.1Claude CodeS = 0.04

  • Serial MZI-v1 · Claude Code · Claude Opus 5.5

    S = 0.039 · scale bar 100 µm

    Fit
    Opus 5.5Claude CodeS = 0.04

  • Serial MZI-v1 · Muse Code · Muse Wonder

    S = 0.038 · scale bar 100 µm

    Fit
    Muse Wonder (high)Muse CodeS = 0.04

  • Serial MZI-v1 · Claude Code · GLM 5.3 Flash

    Invalid final · S = 0 · scale bar 100 µm

    Fit
    GLM 5.3 Flash (high)Claude CodeInvalid · S = 0

  • Serial MZI-v1 · mini-SWE-agent · GLM 5.3 Flash

    Invalid final · S = 0 · scale bar 100 µm

    Fit
    GLM 5.3 Flash (max)mini-SWE-agentInvalid · S = 0

  • Serial MZI-v1 · Codex · GPT-5.6-sol

    Invalid final · S = 0 · scale bar 100 µm

    Fit
    GPT-5.6 SolCodexInvalid · S = 0

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

    Invalid final · S = 0 · scale bar 100 µm

    Fit
    GPT-6.1 Solmini-SWE-agentInvalid · S = 0

  • Serial MZI-v1 · Claude Code · Claude Sonnet 5.5

    Invalid final · S = 0 · scale bar 100 µm

    Fit
    Sonnet 5.5 (high)Claude CodeInvalid · S = 0