Prime Intellect launches Model Factory training platform
11/19/2026 — 11/20, 13:17·1 sources (all)·1 reports (all)
Key takeaways
- US AI infrastructure startup Prime Intellect launched Model Factory, a unified platform spanning pretraining, continued training, supervised…
Background & analysis
On November 19, 2026, the US AI Infra startup Prime Intellect announced the launch of Model Factory, a model training platform. According to the company, organizations can use the platform's supported model and compute configurations to update every parameter of an existing model or to train a new model from scratch.
Model Factory brings pretraining, continued training, supervised fine-tuning, reinforcement learning, and model inference together on a single platform. It also lets teams feed the execution records collected after a model is put into production back into later training rounds, so that groups in scientific research, chip design, finance, and other specialized fields can train, deploy, and keep improving models around their own tasks and data. Prime Intellect states that its underlying reinforcement learning framework, prime-rl, supports training MoE models with over one trillion parameters on clusters of more than a thousand GPUs, with training able to scale up to thousands of GPUs. Based on Qwen3-235B-A22B, the MoE training fused kernel Prime Mega MoE reaches 48% model FLOPs utilization (MFU) on H200 GPUs.
Model Factory is currently offered through two service modes, On-demand hosting and Dedicated deployment, with the Dedicated option allowing customers to bring their own training frameworks. The platform is now available at this stage, with no further milestones reported beyond the launch itself.
AI-generated from 1 reports · updated 2 days ago
- Heat index
- 25
- All sources
- 1
- All reports
- 1
- First seen
- 3 days ago
Reports on this story headlines open the original
Prime Intellect, a US AI infrastructure startup, has launched Model Factory, a training platform that unifies pre-training, continued training, supervised fine-tuning, reinforcement learning, and inference. The company says its prime-rl framework supports training trillion-parameter MoE models on clusters of over 1,000 GPUs, with its Prime Mega MoE fusion kernel achieving 48% MFU on H200 GPUs using Qwen3-235B-A22B.
智东西AI score 60
Other stories people are talking about
- Heat index 1143RisingAnthropic Reaches $1.6 Trillion Market Value in First Trading Week After IPO21 sources
- Heat index 1011Anthropic Mathematician Says Claude Formalized the Generalized Riemann Hypothesis21 sources
- Heat index 622Anthropic IPO total raise reaches $93.5 billion after greenshoe exercise20 sources
- Heat index 596RisingThinking Machines Lab Open-Sources Inkling 2 Mini with 32GB Model Card Recommendation10 sources
- Heat index 490RisingSEC probes Nvidia accounting practices as shares post seven-week slide9 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion

