Best LLMs for multilingual software engineering

SWE-bench Multilingual

Tests software fixes across nine programming languages, helping you assess models for supporting client codebases beyond a single language. Results cover 300 tasks and do not measure developer productivity directly.

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

Limited reportingSaturated

Official benchmark

SWE-bench Multilingual 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	96.6%

View all 1 published reports

Claude Fable 5.296.6%

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

τ²-bench

Tests whether a customer-service agent can resolve requests while using tools and following company policies, helping you assess opportunities to reduce manual support work.

OmniDocBench

Tests extracting text, tables and formulas from documents, helping you choose a model for turning client files into usable data with less manual copying.

CursorBench

Tests whether an agent can complete ambiguous, multi-file coding tasks from real Cursor sessions, helping you choose a model for client development work with less step-by-step direction.