{"title":"API100 Frontier LLM Latency & Throughput Benchmark Dataset","version":"2026.09","audit_date":"2026-09-25","author":"API100 Engineering Team","license":"CC-BY-4.0","citation":"API100 Research (2026). Frontier LLM Latency, Throughput & Gateway Routing Telemetry. https://apihundred.com/benchmarks","total_sampled_requests":50000,"testing_regions":["ap-south-1 (Mumbai)","us-east-1 (N. Virginia)","eu-central-1 (Frankfurt)","ap-southeast-1 (Singapore)"],"models":[{"slug":"claude-3-5-sonnet","name":"Claude 3.5 Sonnet","provider":"Anthropic","ttft_p50_ms":290,"ttft_p95_ms":480,"throughput_tokens_per_sec":72.4,"prompt_cost_per_million":3,"completion_cost_per_million":15,"gateway_overhead_ms":38,"success_rate_pct":99.98},{"slug":"gpt-4o","name":"GPT-4o","provider":"OpenAI","ttft_p50_ms":275,"ttft_p95_ms":440,"throughput_tokens_per_sec":84.1,"prompt_cost_per_million":2.5,"completion_cost_per_million":10,"gateway_overhead_ms":35,"success_rate_pct":99.96},{"slug":"deepseek-r1","name":"DeepSeek R1","provider":"DeepSeek","ttft_p50_ms":340,"ttft_p95_ms":590,"throughput_tokens_per_sec":54.8,"prompt_cost_per_million":0.55,"completion_cost_per_million":2.19,"gateway_overhead_ms":41,"success_rate_pct":99.94},{"slug":"gemini-1-5-pro","name":"Gemini 1.5 Pro","provider":"Google","ttft_p50_ms":310,"ttft_p95_ms":520,"throughput_tokens_per_sec":64.2,"prompt_cost_per_million":1.25,"completion_cost_per_million":5,"gateway_overhead_ms":36,"success_rate_pct":99.97},{"slug":"gemini-1-5-flash","name":"Gemini 1.5 Flash","provider":"Google","ttft_p50_ms":195,"ttft_p95_ms":310,"throughput_tokens_per_sec":142,"prompt_cost_per_million":0.075,"completion_cost_per_million":0.3,"gateway_overhead_ms":29,"success_rate_pct":99.99},{"slug":"llama-3-3-70b-instruct","name":"Llama 3.3 70B Instruct","provider":"Meta","ttft_p50_ms":230,"ttft_p95_ms":390,"throughput_tokens_per_sec":96.5,"prompt_cost_per_million":0.4,"completion_cost_per_million":0.4,"gateway_overhead_ms":32,"success_rate_pct":99.97}]}