ModelPriceWatch.com
Last scan 2026-09-24 Models tracked 270 Providers 35 Cheapest paid Granite 4.0 H Micro $0.017/Mtok in Every price links to its source

KAT-Coder-Pro V2.5

by Kwaipilot

Current coding mid tier

Today's price · per 1M tokens

Input

$0.741

per 1M tokens

Output

$2.96

per 1M tokens

Blended

$1.30

blended $/1M — 3:1 weighted input:output

Cached input

$0.148

20% of input — prompt caching

How this price is scoped: USD is FX-derived from StreamLake's RMB list price (¥5 in / ¥1 cached / ¥20 out per 1M) at 6.7476 CNY/USD (2026-08-07); re-checked on FX drift, not a promo.

Source: official Kwaipilot pricing · read MODELPRICEWATCH.COM · 2026-09-24

Price receipt

We read Kwaipilot's own pricing page on and found KAT-Coder-Pro-V2.5 listed with a price on the same row — that read is where the number above comes from. The page prices it in CNY; the USD above is converted from that list at a dated FX rate — see the price note for the figures. We keep the page text we read; its content hash is 6bd7234dff.

Overview

Kwaipilot KAT-Coder-Pro V2.5 — an agentic coding model from Kwaipilot, Kuaishou's coding-model team, sold first-party through StreamLake (Kuaishou's enterprise cloud) on its Wanqing (万擎) model platform, whose pricing table lists it in RMB (¥5 in / ¥1 cached / ¥20 out per 1M tokens, unit stated as 元/百万 tokens); the USD shown is converted from that RMB list at a dated exchange rate (6.7476 CNY/USD, 2026-08-07). StreamLake's standalone KAT-Coder product page, which scored it 65.2% on SWE-bench-Pro, was taken offline by the vendor on 2026-09-18 and now returns 404, so the Wanqing platform page is the surface this row cites. Context window 256,000 with 80,000 max output. The Apache-2.0 KAT-Coder-V2.5-Dev weights on HuggingFace are a separate research variant, not this hosted SKU. Release date is first public catalog availability (OpenRouter, 2026-07-10), not a first-party announcement.

Capabilities

struck through = not supported
Input 1/4
Text Image Audio Video
Output
Text
Features 3/9
Prompt caching Reasoning Coding Fast inference Long context Open weights Multimodal Web search Realtime

Benchmark performance

accuracy % · higher is better
Terminal-Bench 2.1 vendor
60.7%

Every per-benchmark score we hold for this model is reported by its own vendor, so none is counted toward an accuracy average or any ranking.

Source: KAT-Coder-V2.5 Technical Report (arXiv 2607.05471) · updated · See full rankings →

Specifications

Provider
Kwaipilot
Context window
256K tokens
Modality
text
Parameters
Proprietary
Open source
No — proprietary
Released
Status
Current
Last updated
Tags
codingagenticcaching

Availability verified: listed on Kwaipilot's own page