How Open Models Win
Open model inference exponential growth
Open model inference growth (compared to Google!)
Open models are catching up as the foundation of R&D
Open weights trail the frontier by a few months
(~3 months via Artificial Analysis)
Kimi K3.1 Pro is the 4th-best model overall
on the Artificial Analysis Intelligence Index
Open models occupy a slim slice of the price Pareto frontier
Mapping the open model economy
Revenue growth is key to continued training-scaling, and often captured elsewhere
Competing with integration and vast scale
The opening of an opportunity: enterprises want to own their IP and context
A near-term challenge – convergence of business models
Model builders have the resources to compete in inference
Open models are economically fragmented
The rising costs of training a truly frontier model are looming
Labs will start getting pushed out
Is the value in open models fully resting on keeping up?
The task for open models – build the cheapest, most adaptable, private intelligence
GLM 5.2 was a step change for open agents
- Used by researchers across industry to complement Claude/Codex
- Finetunable to specialized workflows
- The first open model to have agentic product-market-fit like Opus 4.5 in December 2025
Win condition 1: Driving the cost of proven LLM workflows to 0
Win condition 1: Driving the cost of proven LLM workflows to 0
Win condition 1: Driving the cost of proven LLM workflows to 0
Win condition 2: Adaptable, private, controlled intelligence
Win condition 2: Adaptable, private, controlled intelligence
Win condition 2: Adaptable, private, controlled intelligence
An economic engine for ownership and cost-saving
Harvey, August 2026: a Kimi K3 base post-trained with Fireworks for legal work, so law firms can own their models
Mercor and SkyRL, September 2026: Qwen3.5-397B-A17B post-trained on expert data, with the script and weights released
Example: Harvey post-trains Kimi K3 for legal work
Harvey, August 2026: a Kimi K3 base post-trained with Fireworks for legal work, so law firms can own their models
Harvey, August 2026: quality-cost frontier on Legal Agent Bench hold-out tasks; Tenet nearly doubles the all-pass rate of its Kimi K3 base at a stable cost
Open models will thrive as compute becomes less abundant
Let’s build the open future together