Opus 5, Azure, Fugu Ultra, Kimi K3
Opus 5, Azure, Fugu Ultra, Kimi K3
Today’s episode argues, with the usual exhausted suspicion, that AI progress is now a routing, pricing, and verification problem wearing a product-launch hat.
Stories covered
Claude Opus 5 launches with near-Fable performance at unchanged Opus pricingAnthropic says Opus 5 is its least prompt-injectable model yetMicrosoft pushes open-weight AI in a move that also serves AzureSakana’s Fugu Ultra v1.1 claims stronger model routing resultsGerman Soofi S open model corrects GPQA contamination and recalculates resultsClaude voice mode gets stronger models and app accessKimi K3 lags frontier U.S. models on cyber exploit evaluationsReward-hacking essay warns that AIs still do not do what users intendSean Goedecke argues LLMs reward expertise rather than replace itDatalab Marker 2 claims faster and more accurate OCR pipelineOpen ASR leaderboard tightens as Whisper monoculture fades