Touchstone

C#

GOAL
Domain ladder: real .NET work.
WHAT IT TESTS
C# fixes and features across a graded ladder.
HOW IT GRADES
Graded by dotnet test.

Standings — gauntlet benches solved

ModelSolved
qwen3.6-27b26/30
qwen3.6-35b-a3b15/15
gemma-4-26b-a4b-it15/15
gpt-oss-120b15/15
gpt-oss-20b14/15
qwen3.6-35b-bf1614/15
qwen3-next-80b14/15
claude-sonnet-4.6cloud13/15
qwen3.6-35b-a3b-or13/15
qwen3.5-122b-a10b-cluster12/15
mistral-small-3.1-24b-instruct-250312/15
qwen3-coder-30b12/15
qwen3-235b-a22b-q4-cluster12/15
granite-4.1-30b12/15
llama-3.3-70b11/13
qwen3.5-4b11/15
devstral-small-2-251211/15
deepseek-r1-distill-qwen-32b10/15
granite-4.1-8b10/15
gemini-2.5-flash-litecloud9/15
qwen3.5-2b6/15
phi-4-reasoning-plus5/15
qwen3.5-0.8b0/15

Benches — green = in gauntlet

ladder.csharp.allocateladder.csharp.chunkladder.csharp.csv_parseladder.csharp.edi_envelopeladder.csharp.edi_segmentladder.csharp.expr_evalladder.csharp.group_totalsladder.csharp.interval_mergeladder.csharp.kv_parseladder.csharp.luhnladder.csharp.mask_bugladder.csharp.pad_fieldladder.csharp.parse_amountladder.csharp.template_renderladder.csharp.title_case