New Voice Standard Bench: 50+ LLMs ranked on real voice tasks Learn more
Voice Standard Bench
Choose the right model for every voice task
Shortlist the right LLM for your transcripts faster. Voice Standard Bench tests every LLM Gateway model on real voice tasks, giving you a starting point for benchmarks of your own.
Qwen3.5 4B Fast
The results
Models by output performance
Every Gateway model, scored on real transcripts. Pick a task and transcript length above to see performance, cost per hour of audio and response time.
Price performance
Cost per hour of audio, USD
Performance score against cost per hour of audio overall on 2-minute transcripts. The band marks the models with the best price performance. Showing 9 of 50+ models.
Cost per hour of audio
Cost per hour of audio, USD
Cost per hour of audio overall on 2-minute transcripts, cheapest first, among the top 8 models by performance.
Response time
Average response time, seconds
Average response time per call on 2-minute transcripts, fastest first, among the top 8 models by performance.
Every model on this page, one API key. LLM Gateway puts 50+ LLMs behind the same key as your Speech-to-Text, at provider rates with 0% markup.
Head to head
Compare the models you are already considering
Pick three Gateway models and see performance, cost per hour of audio and median response time side by side.
PerformancePerformance score, 0 to 100
Cost per hour of audioCost per hour of audio, USD
Median response timeMedian response time, seconds
Check the winner on your own audio. Upload a file in the LLM Gateway playground, run a prompt over the transcript and switch models to compare answers. No code needed.
Run your own
Benchmark with your own data in LLM Gateway
These scores are a starting point. Run the models that look right on your own transcripts before you decide.