MoonshotAI: Kimi Latest

MoonshotAI Kimi Latest functions as a forward-pointing alias on OpenRouter rather than a frozen model snapshot, always routing requests to whatever release is currently newest in the Moonshot AI Kimi family. The alias was created on April 27, 2026 with a very large 1,048,576-token context window, accepting text, image, and video as inputs while emitting text only. Because the alias moves with the family, it inherits the multimodal reasoning and long-context behavior that the latest underlying Kimi release provides, including access via OpenRouter and the same per-million-token pricing of $2.80 for input and $14.00 for output.

For teams that want to stay on the newest Kimi line without manually swapping model identifiers, the alias removes the operational step of tracking each release. The practical sweet spot is long-horizon multimodal work where the very large context window and combined text, image, and video intake matter, and where a single stable endpoint simplifies integration even as the underlying weights advance. Since this slug is a redirect, workloads that require pinning a specific benchmark profile or a fixed model version should target a named Kimi release instead.

Kilo Gateway~moonshotai/kimi-latestkimi

Open provider page Provider docs

Quick Info

Provider: Kilo Gateway
Model key: ~moonshotai/kimi-latest
Release date: Apr 27, 2026
Last updated: Apr 27, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $0.40
Output token cost: $14.00

Limits

Output tokens: 943,718 tokens
Context window: 1,048,576 tokens

Transparent token rates

Compare MoonshotAI: Kimi Latest pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about MoonshotAI: Kimi Latest

Kilo Gateway

CoverageBenchmark

OpenRouter's Kimi K3 listing confirms a July 16, 2026 release of a 2.8T-parameter open-weight multimodal reasoning model from Moonshot AI, targeting complex coding, knowledge work, and long-horizon agentic workflows including repository navigation, tool use, and image/log/test-driven iteration. The model uses Kimi Delt OpenRouter routes K3 across at least 14 hosts (Morph, DeepInfra, DigitalOcean, Baseten, Together, Moonshot AI, Fireworks, Chutes, Fireworks US, Fireworks Fast, Morph Fast, Phala, Modal, Alibaba Cloud Int.) with per-provider input/output/cache pricing, P50 latency, throughput, and uptime, plus partial benchmark fields f

Kilo Gateway

CoverageBenchmark

Moonshot AI's Kimi K3 has launched through the company's API, kimi.com, mobile apps, Kimi Work, and Kimi Code, with OpenRouter also providing immediate access. The model has 2.8 trillion parameters, native vision support, a 1,048,576-token context window, and is priced at $3 per million input tokens and $15 per million output tokens. Moonshot describes K3 as a long-horizon agent model targeting software engineering, knowledge work, and multimodal reasoning workloads. K3 is built on Kimi Delta Attention and Attention Residuals, uses Stable LatentMoE with 896 experts activating 16 per forward pass, and accepts text, images, and video inputs. On the Artificial Analysis Intelligence Index, K3 scored 57 and ranks fourth among 189 models, behind Claude Fable 5 and two GPT-5.6 Sol reasoning settings, and ahead of Claude Opus 4.8 and GLM-5.2. It also leads Arena.ai's Code WebDev leaderboard, with a separate open-weight checkpoint expected in a later release.