Best LLMs for visual software engineering

SWE-bench Multimodal

Tests software issues that include screenshots and visual context, helping you assess which model could connect a client's visible problem to the code changes needed to fix it.

Updated: Nov 9, 2026Data as of: Nov 9, 2026

Limited reportingNot saturated

Official benchmark

SWE-bench Multimodal results. Each row identifies its benchmark version and evaluation track. Scores from different versions are not directly comparable. Best effort from the latest stored report date per model and version.
Model / creator	Version	Reported score
Claude Fable 5.2Anthropic · Proprietary	Anthropic November 2026 / five trials	74.1%

View all 1 published reports

Claude Fable 5.274.1%

Anthropic November 2026 / five trials

effort: max · Mean of five trials; variant as stated in system card. No Verified or version-specific Pro score inferred. · Tools enabled

Anthropic ↗

Sources and original reports ↗

Other benchmarks

Real-World Finance

Compares financial models, audits and other professional finance deliverables, helping you assess AI support for preparing client analysis.

LiveBench: Coding

Tests coding tasks, helping you compare models for implementation work that you would otherwise write or debug yourself.

IFBench

Tests whether a model follows detailed formatting and output instructions, helping you reduce rework when client deliverables must meet a precise brief.