FlyRouter vs Modelflare

One model on two relays: integrity, speed, price, cache and availability, plus intelligence adjusted for degradation and value for money, with advice for the priority you choose.

Model
codex_pro
openai-first-topup

Which to choose

Five dimensions

Further out is better. Integrity is the share of intelligence kept; the other four are places among all relays of this model.
IntegrityNot checkedSpeedNot checkedPriceCacheNot checkedAvailabilityNot checked
FlyRouterModelflare

What matters most to you

For Balanced, choose Modelflare

Main reasons

Integrity25%

Integrity: no data for at least one relay

Availability20%

Availability: no data for at least one relay

Value20%

Modelflare gives 72.4 points per unit of money, 1582% more than FlyRouter at 4.30

Speed15%

Speed: no data for at least one relay

Price15%

Modelflare costs ¥0.45 per session, 94% less than FlyRouter at ¥7.50

Cache5%

Cache: no data for at least one relay

FlyRouter

Stronger
No clear gap
Weaker
Price¥7.50Value4.30

Modelflare

Stronger
Price¥0.45Value72.4
Weaker
No clear gap

Key figures

Adjusted intelligence

Higher is better

Cost per coding session

2M input + 40k output tokens · Lower is better

Value for money

Intelligence points per ¥1 · Higher is better

All figures

Intelligence

Adjusted intelligence32.332.3
Intelligence keptNot checked100%
Question accuracyNot checked97% (90 questions)

Cost

Value for money4.3072.4
Cost per coding session¥7.50¥0.45
Input price¥3.35¥0.1992
Output price¥20.1¥1.1952
Cache discountNot providedNot provided

Performance

Time to first token—1.04 s
Availability, last 24 hours—97.9%
Samples—192

Route

Groupcodex_proopenai-first-topup
Data securityLevel 0Level 4
Adjusted intelligence = the model's published intelligence index × intelligence kept. Intelligence kept = the relay's question accuracy ÷ the highest accuracy among checked relays of the same model; under 10 questions, no adjustment. Value = adjusted intelligence ÷ cost per coding session. Gaps under a threshold count as close. Advice weighs the metrics by the chosen priority and is a guide only.