Section
Models
The latest Models news, analysed daily by the i10X newsroom.
FilterMoonshot AI
-
Kimi K3 Tops New Geological Reasoning Benchmark
Moonshot AI’s Kimi K3 leads a new geoscience benchmark, showing specialized AI potential for mining and energy. Discover the claims, impacts, and need for transparency in domain-specific models.
By
-
Kimi K2.5: Open-Source Multimodal Model for Visual Coding
Moonshot AI releases Kimi K2.5, an open-weights model for visual coding, tool use, and agent-swarm orchestration. Discover infrastructure…
-
Moonshot AI Open-Sources MoonEP for MoE Training
Moonshot AI releases MoonEP, an MIT-licensed expert parallelism library that reduces network bottlenecks in MoE model training. Optimize…
-
Moonshot AI Kimi K3: Million-Token Context and Open Weights
Moonshot AI’s Kimi K3 delivers 1M-token context and open weights at aggressive pricing. Discover how it disrupts RAG…
-
Kimi K3: Moonshot AI Builds Frontier LLM With Limited Hardware
Moonshot AI’s Kimi K3 delivers strong reasoning, coding, and ultra-long context under hardware limits. It gives Chinese enterprises…
More stories
-
Kimi AI: Moonshot’s Long-Context LLM Efficiency Edge
Moonshot AI’s Kimi pushes 2M+ token context windows with compute-efficient performance on restricted hardware. See how it challenges Gemini and Claude while reshaping enterprise LLM economics.…
-
Kimi K3 Analysis: Performance Claims vs Western AI Rivals
Kimi K3 has generated buzz for claiming to outperform models like GPT-4o and Claude 3.5. This analysis reviews the gaps in benchmarks, pricing, and bilingual capabilities…
-
Kimi K2.7-Code: Internal Benchmarks and Transparency Gaps
Moonshot AI launched Kimi K2.7-Code with a 21.8% internal improvement claim, but lacks comparisons on public benchmarks. Explore the implications for enterprise coding AI adoption.
-
FlashKDA: Moonshot AI’s Open-Source CUDA Kernels for LLM Speed
Discover Moonshot AI’s FlashKDA, a high-performance open-source toolkit using CUTLASS for accelerating Kimi Delta Attention. Achieve faster LLM decoding with variable-length batching, reducing inference costs. Explore…
-
PrfaaS: Global KV Cache for Efficient LLM Serving
Discover PrfaaS, a groundbreaking architecture from Moonshot AI and Tsinghua University that distributes the KV Cache across datacenters worldwide, reducing costs and latency for global AI…
-
Kimi K2.5: Moonshot AI’s Open-Source Visual Agentic Intelligence
Explore Moonshot AI’s Kimi K2.5, an open-source multimodal model featuring Mixture-of-Experts, native vision encoding, and Agent Swarm for efficient multi-agent tasks. Discover its potential to revolutionize…
-
Kimi K2 Thinking: Mastering Long-Horizon Agentic AI
Discover Moonshot AI’s Kimi K2 Thinking, an open reasoning model excelling in 300 sequential tool calls for agentic workflows. Explore its MoE architecture, 256k context, and…
Release tracker
All labs · last 90 days · 16 releases
Gemini 4 Argon
ai.google.dev (source, opens in a new tab)Claude Haiku 5.5
support.claude.com (source, opens in a new tab)Mistral Large 4
docs.mistral.ai (source, opens in a new tab)GPT-6.1 Sol
developers.openai.com (source, opens in a new tab)Claude Sonnet 5.5
support.claude.com (source, opens in a new tab)Claude Opus 5.5
anthropic.com (source, opens in a new tab)GPT-6 Sol & Luna
developers.openai.com (source, opens in a new tab)Grok 4.7
docs.x.ai (source, opens in a new tab)
Show all 16 releases
DeepSeek-V4.1-Flash
api-docs.deepseek.com (source, opens in a new tab)GPT-6 Astra
openai.com (source, opens in a new tab)Gemini 3.8 Flash
ai.google.dev (source, opens in a new tab)Claude Fable 5.1 & Mythos 5.1
support.claude.com (source, opens in a new tab)DeepSeek-V4-Pro (GA)
api-docs.deepseek.com (source, opens in a new tab)Gemini 3.7 Flash
ai.google.dev (source, opens in a new tab)Grok 4.6
docs.x.ai (source, opens in a new tab)DeepSeek-V4-Flash
api-docs.deepseek.com (source, opens in a new tab)
Sources: company announcements & release notes.