Kimi K2 0905

Kimi K2 0905 is an updated release of Moonshot AI's Kimi K2 family, built as a mixture-of-experts language model that activates 32 billion parameters out of 1 trillion total. That sparse-activation design lets the model carry a very large overall knowledge base while keeping the compute footprint of a single forward pass closer to a mid-sized model. The 0905 refresh is described by host platforms as an iteration that improves coding capability, extends context length, and strengthens agentic tool use compared with the original Kimi K2.

The headline practical feature is a 262,144-token context window, which makes the model well suited to long-document reasoning, multi-file code analysis, and agent workflows that need to keep large instruction histories, retrieved passages, or tool transcripts in view. Independent inference providers have published deployment and benchmark material around the same model, including a dedicated latency, throughput, and cost study on DeepInfra and on-demand dedicated-GPU hosting on Fireworks AI, indicating that the weights are available to multiple serving stacks rather than only a single vendor.

Jiekou.AImoonshotai/kimi-k2-0905kimi-k2

Open provider page Provider docs

Quick Info

Provider: Jiekou.AI
Model key: moonshotai/kimi-k2-0905
Release date: Jan 1, 2026
Last updated: Jan 1, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $0.60
Output token cost: $2.50

Limits

Output tokens: 262,144 tokens
Context window: 262,144 tokens

Transparent token rates

Compare Kimi K2 0905 pricing

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

Browse this family

Latest news about Kimi K2 0905

NovitaAI

CoverageRelease Notes

Opper AI's Moonshot release tracker confirms Kimi K2 0905 was released on 5 September 2025 with a 262K token context window, listed input pricing of $0.60 per million tokens and output pricing of $2.50 per million tokens. The tracker places Kimi K2 0905 in the Moonshot release sequence between the original Kimi K2 and The tracker attributes an intelligence score of 15 to Kimi K2 0905 within its composite methodology, without detailing the scoring approach. Out of 7 tracked Moonshot releases shown on the page, 5 are noted as running on Opper, giving developers a consolidated view of the Kimi family's pricing and capability progressio

Jiekou.AI

CoverageBenchmark

MoonshotAI's Kimi K2 0905 is documented as a Mixture-of-Experts model with 1 trillion total parameters and 32 billion active parameters (rumoured), featuring a 262K token context window and a release date of September 4, 2025. Benchable reports a 100% success rate across all tracked benchmarks, perfect accuracy in Hall Pricing on the page is listed at $0.6 per 1M prompt tokens and $2.5 per 1M completion tokens, with Moonshot AI's own endpoint shown as the primary provider entry alongside additional routing endpoints. The model supports tools, response format, and structured outputs, aligning with Moonshot's positioning of the 0905 up