Inference
Proximal

FrontierSWE v2

A 34-task expansion of FrontierSWE for ultra-long-horizon engineering and research work that remains far from saturation.

ModelsModels with a result.15
Top resultBest result on this test.65.5%GPT-6 Astra
Top-3 spreadPoints from first to third.9.2 pts
YearYear of release.2026

Result and price

020406080100$1$3$10$30$100RESULTOUTPUT PRICE PER 1M TOKENS · LOG SCALE

Results

15 results
#ModelResultThe score from the source.
01GPT-6 Astra65.5%
02Claude Opus 5.562.3%
03Claude Fable 5.156.3%
04Claude Opus 552.0%
05Claude Fable 547.0%
06GPT-5.6 Sol32.2%
07GLM-5.330.2%
08Grok 4.729.5%
09Kimi K325.9%
10Grok 4.625.3%
11Gemini 3.7 Flash20.3%
12Gemini 3.8 Flash19.6%
13Qwen3.8 Max15.8%
14Muse Spark 1.212.0%
15Inkling4.1%