buildagentic

Tracker

How far apart are model output prices?

Output tokens are the expensive half of an API bill and the prices span two orders of magnitude. Ranked cheapest to dearest, from the figures we sync from OpenRouter.
Every dot is a row in Language and embedding models, with the source that reported it and the moment it was read.12 tracked →
Output price, $ per million tokens
33 days of readings

Claude Opus 5 leads on output price per million tokens with $25, 24.3x the median of $1.

HighestClaude Opus 5 · $25MedianDeepSeek V3 · $1LowestLlama 3.1 8B Instruct · $0.08Tracked12

Dot height = dollars per million output tokens, dot colour = modality. Left to right is rank, cheapest first.

  • text10
  • code1
  • reasoning1
$0.00$6$13$19$25Output price, $ per million tokensLowestHighestLlama 3.1 8B Instruct$0.08DeepSeek V3$1Claude Opus 5$25Llama 3.1 8B Instruct$0.08textSynced from source, just nowPhi-4$0.14textSynced from source, just nowLlama 3.3 70B Instruct$0.32textSynced from source, just nowQwen2.5 72B Instruct$0.40textSynced from source, just nowGPT-4o mini$0.60textSynced from source, just nowQwen2.5 Coder 32B Instruct$1codeSynced from source, just nowDeepSeek V3$1textSynced from source, just nowDeepSeek R1$3reasoningSynced from source, just nowGemini 3.7 Flash$4textSynced from source, just nowClaude Sonnet 5$10textSynced from source, just nowGPT-4o$10textSynced from source, just nowClaude Opus 5$25textSynced from source, just now
Language and embedding models →

History

What has moved since we started watching

33 distinct days of readings so far. Every observation is kept; none is overwritten.

  • Gemini 3.7 Flash

    $4 +$2 over 33d

    Gemini 3.7 Flash: 36 readings, $2 to $4.

  • Llama 3.3 70B Instruct

    $0.32 −$0.39 over 33d

    Llama 3.3 70B Instruct: 37 readings, $0.71 to $0.32.

  • Claude Opus 5

    $25 +$0.00 over 33d

    Claude Opus 5: 37 readings, $25 to $25.

  • Claude Sonnet 5

    $10 +$0.00 over 33d

    Claude Sonnet 5: 37 readings, $10 to $10.

  • DeepSeek R1

    $3 +$0.00 over 33d

    DeepSeek R1: 37 readings, $3 to $3.

  • DeepSeek V3

    $1 +$0.00 over 33d

    DeepSeek V3: 37 readings, $1 to $1.

  • GPT-4o

    $10 +$0.00 over 33d

    GPT-4o: 37 readings, $10 to $10.

  • GPT-4o mini

    $0.60 +$0.00 over 33d

    GPT-4o mini: 37 readings, $0.60 to $0.60.

  • Llama 3.1 8B Instruct

    $0.08 +$0.00 over 33d

    Llama 3.1 8B Instruct: 37 readings, $0.08 to $0.08.

  • Phi-4

    $0.14 +$0.00 over 33d

    Phi-4: 37 readings, $0.14 to $0.14.

  • Qwen2.5 72B Instruct

    $0.40 +$0.00 over 33d

    Qwen2.5 72B Instruct: 37 readings, $0.40 to $0.40.

  • Qwen2.5 Coder 32B Instruct

    $1 +$0.00 over 33d

    Qwen2.5 Coder 32B Instruct: 37 readings, $1 to $1.

More trackers

Other questions