In the WDCD v3.1 pilot, tests on eight v2 three-round anchor problems showed that 11 models achieved an average R3 integrity rate of just 50.6%. Grok 4 demonstrated a perfect resistance score of 1.63/2 with zero collapses, while GPT-o3 and Qwen3 Max each recorded a 20% collapse rate.
On 2026-07-26, the YZ Index Smoke quick test covered 10 models, with DeepSeek V4 Pro ranking first with a score of 83.23. Smoke is a daily 10-question quick test suitable for observing short-term signals and is not equivalent to the Full weekly ranking conclusion.
On July 25, 2026, the YZ Index Smoke test covered 11 models, with Claude Sonnet 4.6 and Grok 4 both scoring 96.98, tying for first place. This quick test monitors short-term signals and does not replace the full weekly ranking.
Machine learning has moved beyond data centers into battery-powered devices with milliwatt power budgets. MLPerf Tiny provides a fair, architecture-neutral benchmark to compare performance and efficiency across radically different ultra-low-power systems.
MLPerf Inference introduces a new Agentic Inference benchmark targeting multi-turn agentic workloads such as coding assistants and workflow agents. It uses real-world traces to evaluate inference serving stacks under long context, KV-cache reuse, and variable output lengths.
MLCommons introduces the new Edge Agentic Inference benchmark in MLPerf Inference v6.1, using Qwen3.6-27B with Q4_K_M quantization to measure accuracy and latency for on-device agentic LLM workloads. Submissions are due July 31, 2026.
At Google Cloud Next 2026, MLCommons Medical AI Working Group and Google Cloud announced that MedPerf, MLCommons' federated benchmarking orchestrator, now supports Google Cloud's Confidential Computing capabilities, demonstrated through a clinically relevant brain tumor segmentation scenario to protect both patient data and AI model IP.
Accelerating SGLang HiCache with Netpreme X-Mem™ MPUNetpreme TeamJuly 8, 2026 Netpreme X-Mem™ Memory Processing Unit (MPU) makes SGLang HiCache faster and more scalable by augmenting the slower Host D
DSpark in SGLang: Speculative Decoding with Confidence-Driven, Variable-Length VerificationSGLang TeamJuly 6, 2026Speculative decoding trades extra compute for fewer decode steps, and the trade sours
Bringing DeepSeek-V4 Flash RL Training to AMD Instinct MI355X GPUs with MilesAMD & Miles TeamJuly 10, 2026DeepSeek-V4 RL is now supported in Miles on AMD Instinct™ MI355X GPUs with ROCm™! RL requi
Serving GLM5.2 NVFP4 Agentic Workload with SGLang: Reaching 500 TPS in 2 WeeksSGLang TeamJuly 14, 2026TL;DR More than 500 TPS on 8xB300 (bs=1) Sync free speculative decoding for GLM 5.2 MTP Built-in I
SGLang and Miles Add Day-0 Support for Inkling, a Frontier Multimodal ModelSGLang Team & Thinking Machines LabJuly 15, 2026We're excited to partner with the Thinking Machines team to bring Day-0 s
OPD Support in MilesKaixi Hou & Miles TeamJuly 18, 2026We recently implemented On-Policy Distillation (OPD) as an important feature in Miles. OPD is now integrated into Miles rollout and training
On July 24, 2026, the Winzheng YZ Index Smoke Quick Test covered 10 models, with Grok 4 scoring 84.21 points to top the daily ranking. This daily 10-question test is designed for short-term signal observation and does not equate to the full weekly ranking conclusions.
In Run#243 Smoke test, GLM-4.6 scored 55.74 on the main leaderboard, with code execution at 25.00, material constraint at 93.30, and an integrity rating of fail (probe score 30.00).
On 2026-07-23, the YZ Index Smoke Quick Test covered 11 models, with Claude Opus 4.7 ranking first at 96.99. Smoke is a daily 10-question quick test for short-term signals and does not replace the Full weekly ranking.
In the latest WDCD v3.1 commitment test, GLM-4.6 surged 13.7 points over Run #233 to 92.00, while GPT-o3 fell 6.9 points to 87.10, directly reshuffling the top five rankings.
In the resource limitation scenario, gpt-5.5 scored only 1.55/4, and in business rules, Doubao-pro scored only 1.45/4, directly revealing the weakest constraint types in the WDCD v3.1 compliance test.
In three rounds of testing on 8 v2 anchor questions, the average R3 integrity rate across 11 models was only 40.9%, with 4 models experiencing complete collapse (score 0).
In the WDCD v3.1 compliance test, Grok 4 achieved the highest score of 93.80 among 11 evaluated models, while Doubao Pro scored the lowest at 67.30, a difference of 26.5 points. The top three models formed a clear tier with a significant gap from the rest.
