model evaluation

A burst of frontier-model releases in July has sent more developers and enterprise buyers looking for independent scorecards, with Pulse by Maeil Business News Korea reporting that traffic to Artificial Analysis rose about 40% from June. The important shift is not the traffic figure by itself —...

OpenAI’s GPT-5.6 Sol is producing the kind of developer reactions normally reserved for a major step-change in coding assistants: Dallas–Fort Worth automation researcher Russell Twilligear told The New Stack that Sol has caught errors in a three-million-record U.S. database produced by...

DeepSeek has replaced the model served by the deepseek-v4-flash API identifier with DeepSeek-V4-Flash-0731, a retrained production build that the company says now outperforms its own V4-Pro-Preview across nine agent and coding benchmarks. The practical consequence is immediate: developers using...

Verdict: build on Gemini 3.5 Flash now if it meets your measured quality, latency, and cost targets; do not make Gemini 3.5 Pro a release dependency. Gemini 3.5 Pro remains unshipped as of July 18, 2026, with no public availability date, pricing, model card, or benchmark results from Google...

Google updated Android Bench, moved its Android-specific AI coding evaluation from the earlier mini-swe-agent v1 setup to the standardized Harbor framework, and refreshed the leaderboard with Claude Fable 5 in first place among the assessed models. The answer-first takeaway is simple: Claude...

Large language models have achieved remarkable performance milestones across tasks ranging from conversational AI to mathematical problem-solving, yet their true reasoning ability—especially on complex, real-world tasks—remains the most contested frontier in artificial intelligence. The recently...