Kimi K3.1 Pro: 3T-class open-weight flagship; full weights due Nov 4
Moonshot’s Kimi announces Kimi K3.1 Pro as its most capable 3T-class open-weight model with a 1M context, plus a Highspeed variant about 2–3× faster. It is live on Kimi apps and the API; full weights are planned for Nov 4, 2026. Vendor benchmarks and cyber-training-data removal claims are attributed to Kimi; API list prices are below.
Kimi (Moonshot AI) introduced Kimi K3.1 Pro as the first model in its K3.1 series and its most capable 3T-class open-weight model to date, with a one-million-token context window and gains in coding, 3D creation, financial research, and other agentic work. Kimi K3.1 Pro Highspeed is said to run about 2–3× as fast. Both are available today on Kimi.ai, mobile apps, Kimi Work, Kimi Code, and the Kimi API; full model weights are scheduled for Nov 4, 2026. Enterprise access is off by default until an admin enables it.
The platform model list describes kimi-k3.1-pro as a 3T-class flagship; predecessor kimi-k3 is called the world’s first open-source 3T-class model (2.8 trillion parameters) with a 1M context. The blog frames K3.1 Pro as a post-trained Kimi K3 with substantially more post-training compute and joint RL across many domains. We do not report active-parameter counts that do not appear on the official pages we captured.
Vendor-reported benchmarks (attribute to Moonshot/Kimi): on Terminal-Bench 4.0, K3.1 Pro scores 65.7%, ahead of the cited GPT-6 Astra 58.2% and just behind Claude Opus 5.5 at 66.4%; on internal Kimi Code Bench 3.0 Long-horizon it scores 61.2% versus K3’s 55.95%. The post also cites BenchCAD, OfficeQA Pro v2, and other in-house numbers. Footnotes say most results used reasoning effort max, temperature 1.0, and top-p 0.95.
On safety, Kimi says it identified and removed cyber-related data from mid-training and RL for the open-weight release while keeping general coding tasks. On internal xdev-bench-hard, a checkpoint without those removals scored 90.5% pass@3 versus 81.0% for final K3.1 Pro; the company says offensive cyber capability remains behind frontier proprietary models public for at least three months. These are vendor claims, not independent reproductions.
API list prices (per million tokens, excluding tax): kimi-k3.1-pro is $2.00 input, $10.00 output, $0.20 cached input, $2.00 for 5-minute cache writes, and $4.00 for 1-hour writes; kimi-k3.1-pro-highspeed is $4.00 / $20.00 / $0.20 / $4.00 / $8.00. Billing is flat pay-as-you-go with no context-length tiers. Flagship access unlocks after a successful top-up of at least $1. Zero Data Retention is available for API customers.
Sources
- Introducing Kimi K3.1 Pro — Kimi / Moonshot AI
- Model List - Kimi API Platform — Kimi API Platform
- Model Inference Pricing Explanation — Kimi API Platform
- Kimi K3.1 Pro - Kimi API Platform — Kimi API Platform