Skip to content

Models & pricing

All prices are USD per 1 M tokens. Live tariffs: luckyapi.chat.

Anthropic — Claude

ModelStrengthIn · officialOut · officialIn · oursOut · ours
Claude Sonnet 4.6Balanced default — most coding agents$3.00$15.00$0.30$1.50
Claude Opus 4.7Flagship reasoning, long sessions$15.00$75.00$1.50$7.50
Claude Haiku 4.5Fast, cheap, great for tools$1.00$5.00$0.10$0.50

Prompt-cache reads on Claude models are billed at the upstream cache discount (≈ 0.1× input price).

OpenAI

ModelStrengthIn · officialOut · officialIn · oursOut · ours
GPT-5Flagship, agent-friendly$10.00$30.00$1.00$3.00
GPT-5 miniFast & cheap$0.50$2.00$0.05$0.20
o4-miniReasoning, cheap$1.10$4.40$0.11$0.44

Google

ModelStrengthIn · officialOut · officialIn · oursOut · ours
Gemini 2.5 Pro1 M context, vision$1.25$10.00$0.13$1.00
Gemini 2.5 FlashFast multimodal$0.30$2.50$0.03$0.25

Open source

DeepSeek and Qwen runs on our infrastructure — already cheap upstream, but we still pass volume savings through.

ModelLicenseIn · oursOut · ours
DeepSeek V3.2MIT$0.03$0.11
DeepSeek R1MIT$0.05$0.22
Qwen 3 Coder 480BApache 2.0$0.06$0.24
Llama 4 MaverickLlama 4$0.04$0.16

Picking a model for coding

TaskRecommendedNotes
Day-to-day editing in Claude CodeSonnet 4.6Best speed/quality tradeoff
Hard refactors, architecture, debuggingOpus 4.7Worth the 5× cost
Tool calls, fast loops, agentsHaiku 4.590% of Sonnet quality at 10% cost
Long-context analysis (1M tokens)Gemini 2.5 ProBest for whole-repo reads
Budget / personal projectsDeepSeek V3.2Surprisingly capable, dirt cheap

How billing works

  • Pay-as-you-go. No subscription, no monthly minimum.
  • Top up any amount ≥ $5. Unused credit doesn't expire.
  • Each request is billed at completion using the upstream provider's reported token counts.
  • Run /cost inside Claude Code for a live session total.
  • Full history & exports at luckyapi.chat.

Independent. Not affiliated with Anthropic or OpenAI.