SkyClaw-v1.0: A Million-Context Agent Model at Ultra-Low Cost
A high-performance agent model for complex tool use, multi-turn workflows, and real-world task execution. Use the flagship model for stronger results, or switch to SkyClaw-v1.0-lite when speed and cost matter most.
Free trial available now. Need evidence first? See benchmark results.
SkyClaw-v1.0 outperforms Minimax 2.7, DeepSeek V4 Flash, and Qwen 3.6 35B A3B / 27B models across all major agent benchmarks. On OpenClaw-related tasks, its performance approaches that of much larger closed-source models, including DeepSeek V4 Pro, Claude Opus 4.6, and Qwen 3.6 Plus.
In addition to SkyClaw-v1.0, we also offer SkyClaw-v1.0-lite, a much faster and cheaper model that inherits strong agentic performance (e.g., better performance compared to Minimax 2.7). This model is suitable for basic agentic tasks that are more sensitive to costs.
We are launching a free trial period — try SkyClaw-v1.0 or try SkyClaw-v1.0-lite. Following the trial, we will progressively open-source each model version.
Strong Performance on Agent Benchmarks
Across mainstream agent benchmarks and our internally developed Claw task evaluations — including PinchBench, Claw-Eval (with the ^3 stability test), and Skywork-Claw-Bench (Skywork's in-house agent evaluation suite built on the OpenClaw environment) — both the main and lite versions outperform Minimax 2.7, DeepSeek V4 Flash, and the Qwen 3.6 35B A3B and 27B models. Substantial improvements are also observed on related code-task evaluation metrics.
Model Training
SkyClaw-v1.0 was trained around practical agent behavior: complex tool environments, filtered synthetic trajectories, and end-to-end reinforcement learning for more stable multi-step execution.
Agent Environment
Training tasks were built from OpenClaw-style workflows with realistic tool relationships and multi-step user demands.
Synthetic SFT Data
High-quality task trajectories were filtered for both final-answer correctness and the quality of intermediate actions.
Agentic RL
End-to-end reinforcement learning improves generalization, stability, and robustness across agent harnesses.
Pricing and Availability
SkyClaw-v1.0 offers exceptional cost efficiency — pricing is only half or even lower compared to Minimax 2.7 and the Qwen 3.6 series models. A free trial is now available.
| Model | Input (CNY/M tokens) | Output (CNY/M tokens) | Cache Read (CNY/M tokens) | Cache Write (CNY/M tokens) |
|---|---|---|---|---|
| SkyClaw-v1.0 Best Value | 0.5 | 4 | 0.2 | 1.5 |
| SkyClaw-v1.0-lite | 0.3 | 2 | 0.12 | 0.9 |
| DeepSeek V4-Pro | 12 | 24 | 0.1 | — |
| DeepSeek V4-Flash | 1 | 2 | 0.02 | — |
| MiniMax-M2.7 | 2.1 | 8.4 | 0.42 | 2.625 |
| MiniMax-M2.7-highspeed | 4.2 | 16.8 | 0.42 | 2.625 |
Tip: choose SkyClaw-v1.0 for stronger agent performance; choose SkyClaw-v1.0-lite for higher-throughput, cost-sensitive workflows.
API Usage
SkyClaw-v1.0 is available via apifree.ai with an OpenAI-compatible API interface. The lighter model is available from the SkyClaw-v1.0-lite page. Get started in minutes:
cURL
curl https://api.apifree.ai/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $APIFREE_API_KEY" \
-d '{
"model": "skywork-ai/skyclaw-v1",
"messages": [
{"role": "user", "content": "Hello, SkyClaw!"}
]
}'
Python (OpenAI SDK)
from openai import OpenAI
client = OpenAI(
api_key="your-api-key",
base_url="https://api.apifree.ai/v1"
)
response = client.chat.completions.create(
model="skywork-ai/skyclaw-v1",
messages=[
{"role": "user", "content": "Hello, SkyClaw!"}
]
)
print(response.choices[0].message.content)
To call SkyClaw-v1.0-lite, use skywork-ai/skyclaw-v1-lite as the model name.
For full documentation, authentication setup, and advanced usage (tool calling, multi-turn conversations, streaming), visit the SkyClaw API page on apifree.ai →
Showcase
These examples start from natural-language prompts and are completed inside agent frameworks such as Hermes, Claude Code, and Codex. We strongly recommend using SkyClaw-v1.0 as the model inside an agent workflow rather than as a standalone chat model.
Application UI
Production-style layouts with realistic navigation flows.
Jump to demos →Interactive Web & Games
Playable interfaces, physics simulations, and game logic.
Jump to demos →Research Dashboards
Data collection, market reports, and visual analysis.
Jump to demos →Apps
For app-building tasks, we recommend running SkyClaw-v1.0 inside agent frameworks like Hermes, Claude Code, or Codex so it can plan, edit files, test, and iterate through the full workflow.
Flight & Travel Booking App
Search, result browsing, booking, and itinerary planning views.
Open demo →
Instagram-style Social App
Feed, stories, profile, and social interaction surfaces.
Open demo →
Xiaohongshu-style App
Mobile social commerce feed with discovery tabs, cards, and lifestyle content.
Open demo →Interactive Web & Games
SkyClaw-v1.0 excels at generating fully functional interactive web applications — from physics simulations to complete game logic — with correct rendering, smooth animations, and proper user interaction handling.
Bouncing Balls in Rotating Frame
Physics simulation generated from one natural-language prompt.
Open fullscreen →Bingo Match Game
Kid-friendly game UI with complete interaction flow.
Open fullscreen →2048 Puzzle Game
Playable puzzle game with scoring, tile movement, and responsive layout.
Try live demo →Tetris
Complete game mechanics with falling blocks, rows, and scoring.
Try live demo →Super Mario Platform Game
Side-scrolling platform demo with keyboard controls and game-state handling.
Try live demo →Airplane Battle
Arcade shooter demo with enemy waves, movement, and combat interactions.
Try live demo →Chess Game
Human-vs-computer board game flow with move selection and game-state handling.
Open fullscreen →Texas Hold'em Poker
Card-table demo with betting flow, player state, and turn-based game interaction.
Try live demo →Financial Terminal (CN)
Interactive market terminal with charts, stock lists, keyboard cues, and news layout.
Open demo →Tank Roguelike
Top-down tank battle with roguelike progression, enemy waves, and upgrade systems.
Try live demo →Slay the Spire (杀戮尖塔)
Deck-building roguelike with card combat, relic system, and branching ascension paths.
Try live demo →Deep Research & Data Visualization
Beyond coding, SkyClaw-v1.0 can autonomously research real-world topics, collect data from multiple sources, and synthesize findings into interactive, publication-quality dashboards and reports.
