Pick the jobs you use AI for and what matters most. Get one model for each, with the monthly cost and a ready-made OpenClaw config.
Free tool that recommends one AI model per task (coding, creative writing, research, image analysis, agents and more), optimized for quality, value or speed, with an estimated monthly cost.
Pick one or more.
The strongest model for each task.
Rough numbers are fine. Costs update as you type.
Your setup
$29.55estimated per month
Coding
Claude Opus 5.5
Anthropic · High effort · Intelligence 53.6 · Answers in 41 s
$29.55100 tasks
Also good: MiMo-V2.6-Pro ($2.04) · DeepSeek V4.1 Flash ($3.31)
Costs use a typical job size for each task at each model's everyday reasoning setting, thinking tokens included. Your own prompts and outputs will move the numbers.
// OpenClaw configuration generated by dervity.com/tools/model-selector
// Generated: 2026-09-26
{
agents: {
defaults: {
// Primary model: Claude Opus 5.5 (Anthropic)
model: {
primary: "anthropic/claude-opus-5.5",
fallbacks: ["xiaomi/mimo-v2.6-pro"]
},
models: {
"anthropic/claude-opus-5.5": {}, // Best for: Coding
"xiaomi/mimo-v2.6-pro": {},
"deepseek/deepseek-v4.1-flash": {},
},
// Cheapest model for sub-agent tasks
subagents: {
model: "xiaomi/mimo-v2.6-pro",
maxConcurrent: 2
},
timeoutSeconds: 600,
maxConcurrent: 3
}
}
}Was this tool helpful?
It uses the same picks as our best-model-for pages. Each model is scored on the Artificial Analysis Intelligence Index at its strongest reasoning setting that answers within 45 seconds (creative writing uses EQ-Bench instead). Quality picks the top score, Value the most score per dollar among strong models, and Speed the quickest full answer.
Each task has a typical job size, for example about 45,000 tokens for a coding task and 4,000 for an email. We price that at list rates and scale the output by how much each model actually writes, thinking included, in Artificial Analysis's tests. Long prompts, long outputs and caching will move your real bill.
Yes. The export uses OpenRouter model names that OpenClaw accepts, with your top pick as the primary model and the cheapest pick as the fallback and sub-agent. The Config Generator has more options.
Whenever a major model launches, and at least monthly. The date on each best-model-for page shows the last refresh.
An API gateway that gives you many models from Anthropic, OpenAI, Google, Meta and others through one API key, which makes switching models easy. The selector links each pick to its OpenRouter page.
Everything from model picks to CSV converters, with no account needed.