Skip to content

AI API Pricing, per Provider

Discover a diverse range of AI models, grouped by provider, and choose a provider to drill into detailed API pricing for their models.

OpenAI 12 models
$0.10 to $10.00 / M input • Starter to Pro+
CheapestGPT-4.1 Nano · $0.10
FlagshipGPT-6 Astra · $10.00
Anthropic (Claude) 11 models
$1.00 to $10.00 / M input • Starter to Pro+
CheapestClaude Haiku 4.5 · $1.00
FlagshipClaude Fable 5.1 · $10.00
Grok (xAI) 6 models
$1.25 to $2.00 / M input • Pro
CheapestGrok 4.3 · $1.25
FlagshipGrok 4.7 · $2.00
Open-weight 5 models
$0.14 to $3.00 / M input • Together, Groq
CheapestDeepSeek V4 Flash · $0.14
PriciestKimi K3 · $3.00
Compare in-app on Pro

Head to head: which model wins your task?

We ran the same job on four providers’ models, extracting action items into JSON. The top scores tied. What actually separated the models was cost and speed.

Quality vs latency, bubble size = cost per run

Insight: Two models tied at 100, one from a frontier lab and one open-weight. DeepSeek matched the top score for a thirteenth of Claude’s cost. Claude answered four times faster. There is no single winner, only a cheapest-that-passes and a fastest-that-passes, and you decide which one you need.

Don't use cost per token to choose a model

The cheapest token is not the cheapest answer, and a low rate does not mean a model is any good at your task(s). What matters most is which model does the job at the quality and reliability you need, at the lowest price.

Quality by model, cheapest to most expensive (illustrative)

The right answer changes all the time

Evaluating today gives you the right answer today. But the model you chose could silently update its weights and start performing worse tomorrow, and a model that failed before could get better and cheaper. The only way to hold high quality at low cost is to measure on a schedule against your own tasks.

The best model today is not the best model forever (illustrative)

What you get on Pro

  • Score any model on your own task, across providers
  • Cost, quality, and latency side by side
  • Pipeline diagnostics that identify actionable improvements
  • Scheduled benchmarks alert you when models regress or drift
  • API access to integrate measurements into your applications

Drive it from your agent, over MCP

  • Let your agent run the comparisons and read the results for you
  • Work from where you already are: your IDE, chat, or console
  • Over 40 tools your agent discovers automatically, no integration code
  • Agentic recipes for model selection, drift checks, and optimization loops

$129/mo $10 for your first week

Then $129/mo. Cancel anytime.

Model counts, cost ranges, and tier labels come from our live registry, the models we support for evaluation rather than each provider's full catalogue, and update as the registry does. The head-to-head is a real gold-standard run on one task, 2026-10-21; it shows how the models did on that task, not a general ranking, which depends on your task and data. The per-provider pages carry the full per-model tables and their own scored runs.