3Simon Willison's Weblog·
Simon Willison quotes Kimi K3
Simon Willison shares a brief quote or observation about Kimi K3, a model from Moonshot AI. The content body is empty, suggesting this is a short link or quote post. The item signals community attention to the Kimi K3 model release or capability.
Related guides (2)
Related events (8)
Simon Willison on Kimi K3 and the pelican benchmark as a model evaluation tool
Simon Willison writes about Kimi K3, a new model from Moonshot AI, using his informal 'pelican benchmark' as a lens for evaluation. The post reflects on what idiosyncratic, qualitative benchmarks can still reveal about model behavior that formal evals miss. As a tier-2 commentary piece, it offers practitioner-level perspective on a new open-weights or API-accessible model.
Zvi Mowshowitz analyzes Kimi K3 capabilities and benchmark performance
Zvi Mowshowitz (Don't Worry About the Vase) publishes commentary on Kimi K3, characterizing it as a high-performing model with strong benchmark results. The piece appears to be a capability analysis and broader discussion of the model's implications. As a tier-2 commentary source, this provides secondary analysis of a notable model release from Moonshot AI.
Interconnects analysis: Kimi K3 and the open-weights escalation
Interconnects (Nathan Lambert) publishes commentary on Kimi K3, framing it as a significant escalation in the open-weights AI competition with global ecosystem implications. The piece analyzes what Moonshot AI's Kimi K3 release means for the broader open-weights landscape. The body is sparse in the ingested form, but the framing suggests substantive strategic analysis of a notable open-weights release.
Kimi K3 2.8T-A50B released: largest open model ever, Opus 4.8-class performance at Sonnet 5 pricing
Moonshot AI has released Kimi K3, a 2.8 trillion total parameter MoE model with 50 billion active parameters, described as the largest open model ever released. The model is reported to achieve performance comparable to Claude Opus 4.8 while being priced at the level of Sonnet 5, representing a significant cost-performance advance. This release continues a strong week for open-weights models and raises the ceiling for publicly available model scale.
Quoting Mitchell Hashimoto
Simon Willison quotes Mitchell Hashimoto in a brief commentary post. The body content is empty, so the specific substance of the quote is unavailable from this record. Given the source and context, it likely relates to AI tooling or developer workflows given Hashimoto's recent work on AI-assisted development.
Simon Willison quotes Andrej Karpathy
Simon Willison's blog features a quote from Andrej Karpathy, though the body content is not available for review. Given the source and the individuals involved, this likely captures a notable observation from Karpathy on AI/ML topics. Karpathy is a prominent voice in the field whose commentary frequently carries signal for practitioners.
Import AI 465: Open vs closed model gaps, Kimi K3, and Demis Hassabis policy vision
Import AI issue 465 covers the evolving gap between open and closed AI models, the release of Kimi K3 from Moonshot AI, and a policy plan articulated by Demis Hassabis. The newsletter is a curated weekly digest from Jack Clark, a well-regarded voice in AI safety and policy. The framing around open vs. closed gaps and frontier lab policy positioning makes this relevant for tracking competitive dynamics and regulatory developments.
Kimi K2.6: Moonshot AI's 1T-Parameter Vision-Language Model Matches Open-Weights Peers, Trails Top Closed Models
Moonshot AI released Kimi K2.6, a 1 trillion-parameter mixture-of-experts vision-language model with 32B active parameters, designed for long-horizon autonomous coding sessions lasting multiple days and multi-agent orchestration scaling to 300 parallel subagents executing up to 4,000 steps. The model matches Qwen3.6 Max Preview and DeepSeek-V4-Pro on the Artificial Analysis Intelligence Index (scoring 54 vs. their 52) while trailing closed models like GPT-5.5 and Claude Opus 4.7. Weights are freely downloadable from Hugging Face under a modified MIT license permitting commercial use, with API access priced at $0.95/$0.16/$4.00 per million input/cached/output tokens. Notable features include a 256K token context window, native INT4 quantization, a 'preserve thinking' mode for multi-turn reasoning continuity, and a research preview 'claw groups' feature enabling cross-developer agent collaboration.