WhatLLM's review confirms Moonshot AI released Kimi K3 as a hosted service on July 16, 2026, with 2.8 trillion total parameters making it the first announced open model in the three-trillion-parameter class. As a sparse MoE, only 16 of 896 routed experts activate per token, alongside native vision, a one-million-token API pricing is listed at $3 per million uncached input tokens, $0.30 per million cached input tokens, and $15 per million output tokens. The review notes K3 placed #1 in Arena's blind frontend coding ranking at launch and is positioned as a model for agents handling multi-step work like navigating large codebases, oper

Beam AI's model directory entry confirms Moonshot AI released Kimi K3 on July 16, 2026, as a 2.8-trillion-parameter Mixture-of-Experts model with native vision and a one-million-token context window, with only 16 of 896 experts active per token. Architectural features include Kimi Delta Attention and Attention Residual The entry cites independent benchmark results: an Artificial Analysis Intelligence Index score of 57, a GDPval-AA v2 Elo of 1668 (above GPT-5.5 and Claude Opus 4.8, behind Claude Fable 5), a leading AutomationBench-AA score of 53%, and second place on AA-Briefcase. Artificial Analysis estimated roughly $0.94 per comple

On July 16, 2026, Moonshot AI released Kimi K3, a 2.8-trillion-parameter Mixture-of-Experts model, with weights scheduled for release on July 27, 2026, according to analyst Nathan Lambert's Interconnects commentary. The article reports K3 placed #2 overall on the Vals AI index, #3 on the Artificial Analysis Intelligenc The piece situates K3 within the broader open-weights versus closed-models ecosystem, arguing that either the open-to-closed or American-to-Chinese performance gap has narrowed to roughly 3–5 months from a previously debated 6–9 months. Lambert emphasizes that Moonshot is competing with Anthropic and OpenAI with far fe

BenchLM confirms Moonshot published Kimi K3's full 2.8-trillion-parameter weights on July 27, 2026, eleven days after launch, with the Hugging Face repository holding 96 Safetensors shards, the Kimi K3 License, and the technical report. The model is described as a Mixture-of-Experts with 2.8T total parameters and 104B The release uses quantization-aware MXFP4 weights with MXFP8 activations, with vLLM, SGLang, and TokenSpeed listed as supported inference paths. Moonshot recommends at least 64 accelerators for serving, making self-hosting impractical for most teams. API pricing is confirmed at $3.00 per million cache-miss input tokens

Fortune reported on July 16, 2026 that Moonshot AI unveiled Kimi K3, described as 2.7 trillion parameters and the largest open-weight large language model available, further shrinking the performance gap between Chinese and U.S. models. Moonshot's press release called K3 its most powerful open-source coding model to da In its release, Moonshot claimed K3 performed competitively with Anthropic's Fable 5, which the article identifies as currently the most advanced widely available AI model. The piece frames the launch as occurring amid global businesses increasingly questioning the cost of deploying models from Anthropic and OpenAI, po