Qwen 2.5 72B Instruct

Alibaba · text · Open-source · GPU Tier A

liveContext 128K

Requests 30d+14.2%

897K

Tokens 30d+18.4%

2.2B

Revenue 30d+20.2%

$786.6

Gross margin+1.2%

48%

Revenue − node payouts

Active users+9.4%

209

Active nodes+3.4%

1,133

Input $/1M

$0.3

Output $/1M

$0.5

p50 latency

420ms

p95 latency

1.3s

Error rate

2.9%

Usage over time (30d)

Pricing simulator

Elasticity-based what-if · Qwen 2.5 72B Instruct

Input price ($/1M)Output price ($/1M)

Projected revenue 30d

$786.6

0.0%

Projected requests

897k

0.0%

Projected margin

58%

Model assumes elasticity of -1.2 on total price. Replace with regression-derived elasticity per model when telemetry is wired.

p95 latency over time (30d)

Price comparison

$/1M tokens

FAR AI

Input

$0.3

Output

$0.5

Together AI

In: $1.2 · Out: $1.2

FAR saves

−67%

Fireworks AI

In: $0.9 · Out: $0.9

FAR saves

−56%

Revenue economics

How this model generates revenue

Developer pays

$0.3 in + $0.5 out per 1M tokens

Node operator earns

44% of revenue distributed as payout · Tier A

FAR AI gross margin

48%