API Status

fireworks Provider Benchmarks

Comprehensive performance summary covering 7 models.

This provider hub highlights throughput and latency trends across every fireworks model monitored by LLM Benchmarks. Use it to compare hosting tiers, track regressions, and discover the fastest variants in the catalogue.

Visit fireworks Official Website

Provider Snapshot

Models Tracked

7

Avg Tokens / Second

36.66

Avg Time to First Token (ms)

102.86

Last Updated

May 20, 2026

Key Takeaways

  • 7 fireworks models are actively benchmarked with 512 total measurements across 464 benchmark runs.

  • kimi-k2p6 leads the fleet with 47.10 tokens/second, while glm-5 delivers 32.20 tok/s.

  • Performance varies by 46.3% across the fireworks model lineup, indicating diverse optimization strategies for different use cases.

  • Avg time to first token across the fleet is 102.86 ms, showing excellent responsiveness for interactive applications.

  • The fireworks model fleet shows consistent performance characteristics (22.6% variation coefficient), indicating standardized infrastructure.

Fastest Models

ProviderModelAvg Toks/SecMinMaxAvg TTF (ms)
fireworkskimi-k2p647.104.4878.600.00
fireworksminimax-m2p745.202.61100.000.00
fireworksllama-3.3-70b40.105.5769.20720.00
fireworkskimi-k2p537.101.1170.000.00
fireworksdeepseek-v4-pro34.4014.5069.500.00
fireworksglm-532.2017.4048.800.00

All Models

Complete list of all fireworks models tracked in the benchmark system. Click any model name to view detailed performance data.

ProviderModelAvg Toks/SecMinMaxAvg TTF (ms)
fireworksdeepseek-v4-pro34.4014.5069.500.00
fireworksglm-532.2017.4048.800.00
fireworksglm-5p120.508.3837.000.00
fireworkskimi-k2p537.101.1170.000.00
fireworkskimi-k2p647.104.4878.600.00
fireworksllama-3.3-70b40.105.5769.20720.00
fireworksminimax-m2p745.202.61100.000.00

Featured Models

Frequently Asked Questions

Based on recent tests, kimi-k2p6 shows the highest average throughput among tracked fireworks models.

This provider summary aggregates 512 individual prompts measured across 464 monitoring runs over the past month.