Kimi K3.1 Pro replaces Kimi K3 on TopxAI
· kimi, models, api, topxai
Move to kimi-k3.1-pro: 1,048,576-token context, always-on reasoning. Since November 8: a private deployment at $0.50 in / $5 out per 1M tokens.
Update, 2026-11-08: Kimi K3.1 Pro now runs on a private deployment rather than Moonshot's own API, at fixed prices of $0.50 input, $0.30 cache read and $5.00 output per million tokens, with no cache-write charge, and low is the default reasoning effort. A costlier standby deployment can answer, rarely, when the main one fails; such a request is billed at the backup price ($2.40 input, $0.24 cache read, $12.00 output, $2.40 cache write). See the changelog. The prices below are those of the October 24 release.
TopxAI now offers kimi-k3.1-pro in place of kimi-k3. The kimi-official route stayed the same, and so did the default selling prices. The new model's official reference was lower, so at launch TopxAI cost 120% of that reference, or 20% more.
Prices at launch (October 24)
USD per million tokens, using the default 5-minute cache-write category:
| Category | TopxAI | Kimi official reference |
|---|---|---|
| Input | $2.40 | $2.00 |
| Cache read | $0.24 | $0.20 |
| Cache write | $2.40 | $2.00 |
| Output | $12.00 | $10.00 |
The reference was checked against Kimi's pricing on October 24, 2026. Each TopxAI rate is 1.2 times the reference; the previous K3 discount no longer describes this model. There is no long-context price tier. TopxAI offers the default cache-write category only. See the current price page for the rates offered to your account.
Administrator price overrides and deliberate deletions carry to the successor unless it already has its own override. Requests already accepted and historical charges keep their captured prices.
Change the model ID and review the key
Replace kimi-k3 with kimi-k3.1-pro in saved client settings and SDK requests. The old ID returns HTTP 403; there is no forwarding alias. The old pricing URL redirects to the new page.
The base URL remains https://api.topxai.xyz/v1. An unrestricted Kimi or Auto-route key can use the successor. If a key has an allowed-model list, explicitly add the new ID. The upgrade does not broaden restricted keys. Installed clients do not rewrite their saved model settings automatically.
Check reasoning and output settings
Kimi's model guide specifies a 1,048,576-token context, a 131,072-token default output cap and a provider hard ceiling of 1,048,576. TopxAI keeps its published output limit at 131,072 tokens to leave room for the prompt; input and output share the context budget. Reasoning is always enabled; reasoning_effort accepts low, high or max, with low as the default since November 8 (max at launch). Remove settings that try to disable thinking.
Start with the text request in the TopxAI model guide. It also explains image input, the cache-write scope and which endpoint to choose. This release does not claim a measured speed or quality improvement over K3.
Models: kimi-k3.1-pro · Models & pricing
Related
- Kimi K3.1 Pro API pricing
- Kimi K3.1 Pro
- Create an API key and choose a route
- How requests are billed
- Abliterated models on TopxAI: the November 10 prices and how to call them
- Moving to Claude Fable 5.2 on TopxAI
- GPT-6.1 Terra on TopxAI
Other languages: 简体中文