At a Glance
| Vendor | Moonshot AI |
| Family | Kimi K3 series |
| Launched | Moonshot released Kimi K3.1 Pro on October 23, 2026 as a post-trained version of Kimi K3, along with a K3.1 Pro Highspeed variant that it says runs two to three times faster. Weights are promised by November 4, 2026 and were not out as of October 31. |
| Context window | 1,048,576 tokens. The base model is a sparse mixture of experts with 2.8 trillion total parameters and about 104 billion active per token. |
| Pricing | $2 per million input tokens, $0.20 for cache hits, and $10 per million output tokens. Highspeed costs $4 / $20. These figures come from secondary coverage citing Moonshot's platform, and one early report said prices were not yet announced. |
| Access channels | Kimi web and mobile apps, Kimi Work desktop 3.1.0 or later, Kimi Code, and the Kimi API as kimi-k3.1-pro and kimi-k3.1-pro-highspeed. Enterprise admins must switch it on. |
Notable Benchmarks
Moonshot reports 65.7 on Terminal-Bench 4.0 against 66.4 for Claude Opus 5.5 and 58.2 for GPT-6 Astra, but well behind both on Program Bench and Terminal-Bench-Science. Independently, Vals AI ranks it 8th of 27 on the Vals Index at 62.52 percent and 3rd of 55 on Terminal-Bench 4.0. Moonshot deliberately cut cyber data from training, and it says it aims "not to release open weights that would extend the absolute frontier of cyber capability."
Strengths
Close to the leading closed models on terminal coding at a far lower price, with strong office and long-running software results, and an open-weight release to follow.
Limitations
Self-hosting is out of reach for nearly everyone: estimated raw size is 2.8 TB at 8-bit, 1.4 TB at 4-bit, and 700 GB at 2-bit, before cache and buffers. No license has been named for the weights. Many comparison figures are Moonshot's own.
Brand-Visibility Implications
Kimi matters to brands that sell in China and to brands reached by developers using low-cost models in agents. Weights also let others host and tune the model, so answers may vary by provider once they land. Check baselines on Moonshot's own API first. Treat the November 4 date as a promise until the files appear. See Kimi K3 brief and Kimi K3 local hardware requirements.
How Presenc AI Tracks This Model
Presenc AI monitors brand visibility on Moonshot AI's Kimi K3 series as part of continuous multi-platform AI visibility tracking. We sample Moonshot Kimi K3.1 Pro across representative prompt sets daily, compare against competitor performance on the same prompts, and flag material mention-rate changes so brand teams can respond quickly when AI representation shifts.