Frontier Log

Benchmark · InactiveSuperseded by FrontierSWE v2

FrontierSWE, release not stated

Created by Proximal, 16 Apr 2026 · Higher is better · Website

020406080100Apr 2026Jul 202681.2%Kimi K378.1%GLM-5.373.5%Qwen3.8-Max74.4%GLM-5.240.7%Qwen3.7-Max30.5%GLM-5.1020406080100Apr 2026Jul 202681.2%Kimi K378.1%GLM-5.373.5%Qwen3.8-Max74.4%GLM-5.240.7%Qwen3.7-Max

Settings and run setups

ScoringProviders
DominanceAlibabaMoonshot AIZ.ai

Download image

Results

ModelScoreSource
Kimi K316 Jul 202681.2%Kimi K3: Open Frontier IntelligenceChecked A person checked this number against its source on 11 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Kimi Code harness · max reasoning effort · dominance recomputed by Moonshot with the official script, as of Jul 16, 2026Our capture
GLM-5.314 Aug 202678.1%GLM-5.3: Frontier Coding with Emergent Cyber CapabilitiesCorroborated The same number is printed in 1 other lab document: 1Run by Proximal · harness not stated · max effort · 1M context, 128K max output · dominance as of Aug 14, 2026Our capture
GLM-5.216 Jun 202674.4%GLM-5.2: Built for Long-Horizon TasksCorroborated The same number is printed in 1 other lab document: 1Run by Proximal · harness not stated · max effort · 1M context, 128K max output · dominance as of Jun 16, 2026Our capture
GLM-5.216 Jun 202667.5%GLM-5.3: Frontier Coding with Emergent Cyber CapabilitiesCorroborated The same number is printed in 1 other lab document: 1Run by Proximal · harness not stated · dominance as of Aug 14, 2026 · reported later · another report, not chartedOur capture
Qwen3.8-Max3 Aug 202673.5%Qwen3.8-Max: A New Bar for Coding and CoworkChecked A person checked this number against its source on 10 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.Claude Code harness · dominance recomputed by Alibaba with the official script, as of Aug 3, 2026Our capture
Qwen3.7-Max20 May 202640.7%Qwen3.8-Max: A New Bar for Coding and CoworkChecked A person checked this number against its source on 10 Nov 2026. This confirms the number matches what the lab published. It doesn't confirm that the score was reproduced.harness not stated · dominance recomputed by Alibaba with the official script, as of Aug 3, 2026 · reported laterOur capture
GLM-5.17 Apr 202630.5%GLM-5.2: Built for Long-Horizon TasksCorroborated The same number is printed in 1 other lab document: 1Run by Proximal · harness not stated · dominance as of Jun 16, 2026 · reported laterOur capture

All benchmarks