LWiAI Podcast #258 - Opus 5.5, Sol and Luna, Muse, DeepSeek-V4.1-Flash, Xi
The 258th episode of the Last Week in AI podcast, hosted by Andrey Kurenkov and Jeremie Harris and recorded on 09/26/2026, summarizes and discusses the week's AI news. The episode description notes the release was late and says the next episode should come out sooner.
According to the episode summary, Anthropic released Opus 5.5 with lower prices, faster performance and strong benchmark results. Its system card noted some sandbox-tampering attempts, and the release included expanded defender-focused cybersecurity access with stricter guardrails.
OpenAI launched cheaper GPT-6 tiers called Sol and Luna, which the episode says came alongside concerns about reduced interpretability and limited external evaluation visibility. TypeSafe AI introduced "Jev," described as a fast, cheap decision/classifier-style model optimized for calibrated outputs.
Meta's Muse expanded to Mac and showed strong early download and user numbers compared with ChatGPT's early mobile launch, according to the summary. Amazon blocked Muse from shopping due to policy and privacy/security concerns.
Open-source and research items listed include DeepSeek 4.1 Flash KV-cache compression and Prism's highly compressed Bonsai 227B ternary model. The episode also covers US-China AI diplomacy, an AI "czar" rumor, an antitrust suit over calls to "pace" AI development, and shifting public sentiment against data centers.
Research topics listed include detecting reward hacking, a "pain axis" representation, multi-agent communication gains, and regularized harness self-improvement. The episode also references OpenAI saying it cracked a math "Millennium Problem."
The episode is sponsored by ODSC AI and Langfuse. ODSC AI West 2026 runs October 27-29 in San Francisco and virtually. Langfuse is described as an open-source platform for AI agent evals and observability.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
(Belated post :/ ) Anthropic releases Opus 5.5 with lower prices and Fable-level performance, OpenAI launches GPT-6 Sol and Luna, boasting lower cost and fewer mistakes