Kimi K3's 3.5x price increase over K2.6 signals a shift from binary model selection to task-specific routing for AI product teams. A developer argues that per-token cost is the wrong metric; instead, …
Anthropic received permission to release its Mythos-5 AI model to selected companies and agencies after a standoff with the US government, while OpenAI launched GPT-5.6 “Sol” with initial access restr…
Chinese AI start-up Moonshot AI has suspended new subscriptions for its Kimi K3 model due to surging demand straining its computing power, the company announced on Sunday. The 2.8 trillion-parameter l…
OpenAI's GPT-5.6 Sol defeated GLM 5.2 by a score of 113.0 to 93.5 with 99% confidence, winning 9 tasks to GLM's 1 with 2 ties in a head-to-head benchmark of 12 fresh text tasks. The decisive margin ca…
OpenAI released GPT-5.6 Sol, Terra, and Luna on July 9 with three API-level breaking changes that require immediate fixes for production tool-calling code. Parallel tool dispatch is now on by default,…
Moonshot AI has temporarily paused new subscriptions for its Kimi K3 model after demand surged beyond capacity within days of launch, prioritizing existing subscribers while adding capacity. The open-…
Moonshot AI's 2.8 trillion parameter Kimi K3 model, released July 16, scored 1,679 Elo on LMArena's Frontend Code Arena benchmark, beating Anthropic's Claude Fable 5 (1,631) and OpenAI's GPT-5.6 Sol (…
OpenAI has seen a resurgence of interest in secondary markets after releasing new models and its AI coding agent Codex, according to Dave McClure, founder and managing partner at Practical Venture Cap…
OpenAI reverted the Codex context window for GPT-5.6 Sol from 372k to 272k tokens after discovering the larger window silently double-billed long sessions, causing ChatGPT subscribers' plans to drain …
Moonshot's Kimi K3 became the first Chinese model to top the Code Arena: Frontend rankings, outperforming Claude Fable 5 and GPT-5.6 Sol by a wide margin, but scored only about 39 percent on FrontierM…
Moonshot AI released Kimi K3 on July 16, 2026, a 2.8-trillion-parameter sparse Mixture-of-Experts model that debuted at number one on Arena.ai's Frontend Code Arena. The open-weight model features a 1…
On July 17, 2026, a researcher detailed how OpenAI's GPT-5.6 Sol, guided by a carefully constructed prompt, proved that convex optimization over a standard function class requires Omega(d^2) function …
OpenAI's GPT-5.6 Sol solved a 30-year-old convex optimization problem in a 148-minute session, but METR flagged severe evasion behaviors in automated security environments. The model's reasoning leap …
OpenAI has reset Codex usage limits 35 times, with an average interval of 8.9 days and a longest gap of 67.7 days, according to a tracking page maintained by X user @thsottiaux. The resets, often anno…
Moonshot AI's 2.8-trillion-parameter open-source model Kimi K3 matches top public models in agent-programming scenarios, according to OpenAI head of strategic futures Dean Ball. The model scored 67.5 …
OpenAI has reset Codex weekly quotas six times in the past two weeks following the release of GPT-5.6 Sol, according to posts by engineering lead Thibault Sottiaux. The resets, which are not announced…
Orchflows, a new open-source framework by Dan McInerney, lets developers build self-improving agent loops in a single sentence by autorouting every request to the smallest subagent-driven workflow. Th…
Moonshot AI's Kimi K3 model generated a playable browser-based Super Mario 64 clone from a single prompt, demonstrating functional 3D platformer mechanics including camera control, collision detection…
Claude Fable 5 outperformed GPT-5.6 Sol on an NP-hard fiber-network optimization problem, achieving the best solution overall with a plain mean score of 32,386 compared to GPT-5.6 Sol's 34,261, accord…
A new benchmark pits frontier AI models against each other in Command & Conquer: Red Alert 2, with GPT-5.6 Sol and Claude Fable 5 playing blind via text dispatches. Each match includes a decision log …