//the open tab for ai engineers

Test. Compare.Debug any AI model.

AI API testing workbench for Claude, GPT-4o, Gemini, DeepSeek, Llama and Bedrock — browser-only, no backend, BYOK.

Test, compare and debug LLMs from Anthropic · OpenAI · Google · AWS · DeepSeek · Meta in one place.
Bring your own key — never touches our servers.

8core tools
6providers
$0to start
KEYyour keys
[04]AI Ecosystem
curated links
→ Model Providers & Hubs
→ Dev Tools & Agents
→ Orchestration & SDKs
→ Local AI & Inference
→ Vector Databases
→ Observability
→ Consumer AI
[06]Articles
11 articles
PerformanceOptimizing TTFT in Next.js
EconomicsMulti-Provider Cost Analysis
SecurityZero-Backend Architecture
PromptingCoT vs Few-Shot Prompting
AnthropicClaude Extended Thinking
ComparisonGPT-4o vs Claude for Code
DebuggingReading Streaming Metadata
DeepSeekDeepSeek V3 Deep Dive
PerformanceUnderstanding Context Windows
EconomicsPrompt Caching: Claude vs Gemini
SecurityAPI Key Security Setup
→View All
[07]Tools Guide
Manual & Help
Core Tools
Prompt Tools
Catalog
[FAQs]Common Questions

Frequently Asked Questions

How do I test Claude, GPT-4o, and Gemini without a backend?+
AIWorkbench.dev is a browser-only application. You paste your API keys directly into the workbench, and your browser makes HTTPS requests straight to Anthropic, OpenAI, and Google. No proxy, no backend, no key storage on our servers.
What is BYOK and why does it matter for privacy?+
BYOK (Bring Your Own Key) means your API credentials never leave your browser. We do not have a database, a server, or any infrastructure that could log your prompts or keys. Open your Network tab and verify every request goes directly to the provider.
How do I compare LLM models side by side?+
Use the Compare tool. Enter a single prompt, select multiple models, and fire them simultaneously. The workbench streams responses in real time so you can evaluate quality, speed, and token usage head-to-head.
Can I calculate API costs before sending requests?+
Yes. The built-in Cost Calculator uses live provider pricing to estimate your spend based on input and output token counts. It also factors in prompt caching and extended thinking budgets.
Is my data safe when using AIWorkbench.dev?+
Absolutely. There is no server processing your requests, no database storing your history, and no analytics tracking your prompts. Your data flows directly between your browser and the AI provider.