Fireworks and Baseten approach decacorn status as inference startups accelerate; OpenRouter raises $113M Series C

Fireworks raised funding in a round valuing the company at $15B (described as "in talks", a 3.75× increase in 7 months); Baseten is raising at $11B valuation (2.2× in 3 months); OpenRouter announced a $113M Series C led by Capital G, growing from 5T to 25T tokens weekly over 6 months.

The trend reflects a shift in AI from experimentation to production use, with multimodel inference requiring routing infrastructure—OpenRouter's growth was framed as a consequence of this production transition.

In parallel, harness engineering (model + harness + evaluation loop) is emerging as the main differentiator for coding agents over raw model strength alone, with benchmarks like DeepSWE showing stronger separation between top performers than public leaderboards typically display.