OpenAI
GPT Image 2
- Success rate
- —Not published
- P95 generation
- —Not published
- Samples
- —Not published
- Services
- 10
Official priceNo published priceOpen model
AI model / relay evaluation
Folkbench: AI API relay reviews and rankings
Check availability, response speed, and stability before you choose a service
- Find models and services
- Relay evaluation
- Quick model testing
- Public evaluation reports
- Model explainer articles
- Image and video evaluation
- User experience sharing
- Vendor onboarding
03 / IMAGE + VIDEO EVALUATION
See how image and video models perform.
Compare the main signals for media generation: success rate, P95 generation time, published samples, and service coverage
- Models
- 13
- Success rate
- —
- P95 generation
- —
- Service coverage
- 29
Published signals
Google Gemini
Gemini 3.1 Flash Image
- Success rate
- —Not published
- P95 generation
- —Not published
- Samples
- —Not published
- Services
- 2
Official priceNo published priceOpen model
ByteDance
Seedream 5.0 Lite 260128
- Success rate
- —Not published
- P95 generation
- —Not published
- Samples
- —Not published
- Services
- —Not published
Official priceNo published priceOpen model
Success rate and latency appear after a published media snapshot is available.
Public rankings
See the ranking before you choose
Availability, response speed, price, and report status sit on one page. Read the ranking, then open the service you want
Availability, response speed, and price on one page
Each row shows whether a report is available
Open that service after you choose it
ModelGPT 6.1 Sol
- IKunCodeData security 0Availability100.00%Input ¥0.50 / 1M tokens · Output ¥2.50 / 1M tokensView report →
- 智流 FluxLaneData security 4Availability97.92%Input ¥0.38 / 1M tokens · Output ¥1.90 / 1M tokensView report →
- ApiporterData security 0Availability97.92%Input ¥0.60 / 1M tokens · Output ¥3.00 / 1M tokensView report →
- ModelflareData security 4Availability100.00%Input ¥0.20 / 1M tokens · Output ¥1.01 / 1M tokensView report →
- Fuel ApiData security 0Availability100.00%Input ¥0.56 / 1M tokens · Output ¥2.80 / 1M tokensView report →
Fair comparison
The ranking has clear rules
The same model is compared with the same published checks. Each result shows when it was measured
Only services for the model on the board
Every service uses the same published checks
Commercial placement does not set the order
Same model
Services are compared for the model shown at the top of the ranking
Same standard
The ranking uses the same published checks for every service
Date shown
Every result carries the date it was measured or published
Business stays separate
Commercial placement does not decide the order of the ranking
Evaluation reports
Every check ends with a clear report
The report says what was checked, when it was checked, and what came back. Service, time, and result stay together so you can review them
The report says what was checked
It also says when the check happened
Service, time, and result can be read together
FOLKBENCH / REPORT
ModelGPT 6.1 Sol
Was it available?See whether the service responded
How fast was it?See the response time for the completed check
Were there problems?See errors, timeouts, missing data, or other notes
Service test
Run a check when you need more detail
Pick a service, run a check you are allowed to make, and read the result in a report
Start with a model and service from the directory
Confirm you may test it before you begin
The finished check comes back as a report
Current statusOn-demand testing
- Choose a service
Start with a model and service from the directory
- Run the check
Confirm that you are allowed to test the service
- Read the report
Use the result to decide what to try next
Models and services
See what is covered
Browse the models and services already listed in the directory, and see how far the coverage goes by series
Look through listed models by series
See which services are in the directory now
A listing does not mean every report is published
Model series15
- OpenAI15 models
Anthropic12 models
Google Gemini15 models
DeepSeek6 models
Zhipu GLM7 models
Alibaba Qwen17 models- xAI Grok6 models
Kimi4 models
ByteDance10 models
Kling6 models
MiniMax5 models
Tencent Hunyuan2 models
A listing does not mean that every model has a complete published report
Ranking trust
Published rules. Fair comparison
The ranking uses published results and keeps commercial display separate from order
Services are ordered using published evaluation results
A missing result stays visible instead of being guessed
Business relationships do not directly change the ranking
Start with one model
Check the result before you choose
Use the ranking for a quick view, then open the report when you need more detail