Releases

Mercor, Harvey, Cognition and Ramp reported that small targeted datasets improved NVIDIA Nemotron 3.5 Super on held-out legal, software-engineering and accounting work.

Mercor introduced a research team focused on evaluating AI on real work, measuring expert-data quality and post-training models to close observed capability gaps.

Mercor said its infrastructure was running more than 2.5 million model-evaluation workloads each week across 241 models.

Mercor updated APEX-Agents with revised task specifications, tooling, and environments to improve leaderboard accuracy.

Mercor reported that its expert network had grown to more than five million domain experts as it announced the Deeptune acquisition.

Mercor announced an agreement to acquire Deeptune, adding reinforcement-learning environment capabilities and its team.

Mercor reported more than four million vetted experts and more than 1.9 million referrals across its network.

Mercor released the open APEX-Agents benchmark for long-horizon professional work; frontier agents completed fewer than 25% of tasks on the initial evaluation.

Mercor raised a $350 million Series C led by Felicis.

Mercor raised a $100 million Series B led by Felicis.

Mercor raised a $30 million Series A led by Benchmark.

Mercor announced $3.6 million in seed funding led by General Catalyst.

Mercor publicly launched its fully automated platform for assessing and matching talent with companies.