SkyClaw-v1.0: A Free Agent Model Closing In on Opus 4.6
Kunlun Tech releases the SkyClaw-v1.0 Agent model: performance approaching Claude Opus 4.6 at half the price of mainstream models, OpenAI-interface compatible, free for a limited time.
SkyClaw-v1.0: A Free Agent Model Closing In on Opus 4.6
Kunlun Tech releases the SkyClaw-v1.0 Agent model: performance approaching Claude Opus 4.6 at half the price of mainstream models, OpenAI-interface compatible, free for a limited time.
If you run Agent tasks with Claude Code, OpenClaw, or Hermes and wince every time the Token bill arrives, this article is worth two minutes of your time.
Kunlun Tech has just released SkyClaw-v1.0 and its lightweight sibling SkyClaw-v1.0-lite. This is a model designed specifically for Agent scenarios — not a general-purpose model that "does Agents on the side," but one trained from day one with a single goal: "get the task done."
The core selling points: performance approaching Claude Opus 4.6, at half the price of mainstream models — and it is free for a limited time right now.
What Is SkyClaw-v1.0
SkyClaw-v1.0 is an Agent-native model from Kunlun Tech. Unlike most "general model + tool-calling shell" solutions on the market, the model was steeped in real Agent environments during training — tool calling, multi-step task decomposition, and long-chain execution are baked in at the lowest level.
The two versions are positioned as follows:
- SkyClaw-v1.0 (flagship): aimed at enterprise-grade complex tasks — multi-step planning, complex tool-chain invocation, long-context multi-turn interaction
- SkyClaw-v1.0-lite (lightweight): aimed at high-frequency, cost-sensitive scenarios — fast and cheap, ideal for everyday Agent assistance
Core Capabilities
Performance: across multiple Agent benchmarks, SkyClaw-v1.0 beats open-source models like Minimax 2.7 and DeepSeek V4 Flash, and closes in on closed-source leaders like DeepSeek V4 Pro, Claude Opus 4.6, and Qwen 3.6 Plus.
Price comparison:
| Model | Input price | Output price |
|---|---|---|
| SkyClaw-v1.0 | 0.5 yuan/million tokens | 4 yuan/million tokens |
| SkyClaw-v1.0-lite | 0.3 yuan/million tokens | 2 yuan/million tokens |
Compared with Sonnet 4.6, SkyClaw-v1.0's input price is only about 1/43, and its output price about 1/27.
Framework compatibility: deeply adapted to mainstream Agent frameworks like OpenClaw, Claude Code, Hermes, and Nanobot, and compatible with the OpenAI interface format. For developers, switching basically means changing the baseURL and API Key.
Real-World Results
SkyClaw-v1.0 performs impressively in real Agent scenarios. Here are a few typical use cases:
Financial terminal: building a multi-module financial markets dashboard from scratch — scrolling global indices, a news feed, mini candlestick charts for watchlist stocks — SkyClaw-v1.0 finished it quickly.
AI industry weekly report system: automatically scraping data from RSS, GitHub, HuggingFace and other sources, cleaning and classifying it, running trend analysis, and finally generating an interactive weekly report page. When it finished, it even suggested setting the whole thing up as a scheduled task.
How to Get Started
There are two ways to get started:
Option 1: The Skywork platform (zero config)
Visit tiangong.cn directly, pick the SkyClaw-v1.0 model, and start using it — nothing to install.
Option 2: API calls (recommended for developers)
- Register an APIFree account (apifree.ai) and get an API Key
- The API is OpenAI-format compatible, supporting streaming output, tool calling, and multi-turn conversation
- For developers already using the OpenAI interface, you basically just swap the baseURL and model name
API documentation:
- SkyClaw-v1.0: https://www.apifree.ai/model/skywork-ai/skyclaw-v1?tab=api
- SkyClaw-v1.0-lite: https://www.apifree.ai/model/skywork-ai/skyclaw-v1-lite?tab=api
Tip: both models are currently free for a limited time; after the trial period ends, Kunlun Tech plans to progressively open-source each version of the model.
Who It Fits
- Individual developers: if you want top-tier Agent capability on a limited budget, SkyClaw-v1.0's free period is the best trial window
- Small and mid-size teams: invoking Agents at high frequency for automation — the lightweight version can slash Token costs
- Agent framework users: if you're on OpenClaw, Claude Code, or Hermes, you can switch the underlying model directly and test the results
Project homepage: https://skyworkai.github.io/skyclaw/
Toolin Editorial Team
Categories
Related articles
Baidu DuMate Hands-On Guide: A Homegrown Codex That Lets You Run Office Work by Voice
From installation to automation, a complete breakdown of Baidu DuMate's request -> authorize -> execute -> deliver pipeline: cross-app work, scheduled tasks, and the credits bill, all explained in one article.
DeNovoSWE: The First Long-Horizon Doc2Repo Training Set That Teaches Code Agents to Build Repositories
Renmin University of China releases an open training set of 4818 real task instances targeting repository-level code generation, lifting Qwen3-30B's pass rate from 5.8% to 47.2%.
Google Workspace CLI: One Command Lets AI Agents Take Over Your Mail, Drive, and Calendar
An open-source CLI with nearly 30K stars wraps Gmail, Drive, Calendar, and Sheets into one unified interface, ships 100+ Agent Skills built in, and works for humans and AI alike.
Deemos Hyper3D Rodin Gen-2.5: 3D Generation Enters Its Thinking Era
The world's first 10M-polygon-class 3D generation model: 1M polygons in 4 seconds, 12K high-res textures, five switchable thinking depths, with B2B revenue exceeding the rest of its niche combined.
Xiaomi MiMo-V2.5-Pro-UltraSpeed: A 1T-Parameter Model Running at 1000 tokens/s
Xiaomi's flagship ultra-speed inference model breaks 1000 tokens/s of output on a standard 8-GPU commodity node, with output priced at 18 RMB per million tokens.
Google Ships Two Lightweight Creative Models: Images in 4 Seconds, Video in 10
Google releases the Nano Banana 2 Lite image model and the Gemini Omni Flash video model, built for speed and savings — $0.034 per image and $0.10 per second of video, already embedded in the Gemini App and YouTube Shorts for free.