Overview
San Mateo–based open-model-inference and training platform for specialized-intelligence — fine-tuning and serving open/custom models. Founded 2022 by lin-qiao (ex-Meta) and co-founders; Nvidia-backed.
Funding
- 2026-07-16: 17.5B** valuation — Atreides, Index, TCV lead; nvidia, Lightspeed, Evantic, Bessemer, Menlo participate (2026-07-17-fireworks-series-d-businesswire)
- Prior: Series C Oct 2025 ~4B (per Sacra secondary reporting)
Ember-1 Research Model (September 2026)
- 2026-09-23: Released ember-1 — kimi-k3 derivative with ~40% fewer reasoning tokens at comparable benchmark quality (2026-09-27-fireworks-ember-1-efficient-reasoning)
- Research preview on serverless; 50+ training experiments, 200+ evaluations
- SWE-bench Verified 92.2%; Terminal Bench 2.1 82.0% (vendor-reported)
- One enterprise customer in production after A/B tests
Benchmarks vendor-reported; not independently reproduced.
Scale Claims (company-reported)
-
$1B annualized revenue run rate (5× YoY)
-
40 trillion tokens/day; >95% specialized models
- ~200 employees; target ~600 by end-2026
- Historically ~50% revenue from cursor; diversified to Elastic, GitLab, MongoDB, Uber, Shopify
Related
- lin-qiao
- atreides-management
- open-model-inference
- specialized-intelligence
- nvidia
- cursor
- ai-investment-trends
- open-weight-models
- ember-1
- reasoning-token-optimization
- inference-cost-optimization
- kimi-k3