Perplexity has post-trained a GLM 5.2 preview model that it says can match Claude Opus 4.8 while costing about one-third as much. The company says the model uses a cheaper open-source base with guidance from a frontier advisor and is already running in production.
Perplexity has post-trained a GLM 5.2 preview model that it says can match Claude Opus 4.8 while costing about one-third as much. The company says the model uses a cheaper open-source base with guidance from a frontier advisor and is already running in production.
Perplexity has fine-tuned a Chinese open-source AI model, GLM 5.2 preview, and says the result matches Claude Opus 4.8 at roughly one-third of the cost. According to the company, the post-trained model combines a lower-cost open-source foundation with a frontier advisor and is already live in production.
Perplexity has fine-tuned a Chinese open-source AI model, GLM 5.2 preview, and says the result matches Claude Opus 4.8 at roughly one-third of the cost. According to the company, the post-trained model combines a lower-cost open-source foundation with a frontier advisor and is already live in production.
The development matters because it points to a practical path for companies trying to deploy advanced AI without relying only on the most expensive frontier models. If a cheaper base model can be adapted to perform near premium systems in production, businesses may have more flexibility in how they manage AI costs and infrastructure.
For crypto and fintech firms, where teams often use AI for research, customer support, developer tooling, compliance workflows, and market monitoring, lower-cost model performance could influence how products are built and scaled. The source does not indicate any direct crypto integration, but the broader cost-performance shift is relevant to technology-heavy sectors.
Perplexity's approach also reflects a wider trend in AI: pairing open-source models with post-training methods to narrow the gap with closed, high-end systems. The key claim is not simply that GLM 5.2 preview is cheaper, but that Perplexity has already put the fine-tuned version into production use.
The company has not framed the release as a consumer product announcement in the supplied material. Instead, it presents the model as an operational deployment that could show how frontier-level AI capabilities may become more economical through targeted fine-tuning.
Keep exploring
Bitget plans to exit the Japanese market and close all remaining positions by the end of the year. The move adds to ongoing shifts among crypto companies navigating Japan’s regulatory environment.
Read
CoinDesk’s State of Crypto coverage counts down recent developments shaping the crypto policy and market landscape. The report frames these changes as important for how companies and participants navigate the sector.
Read
Strategy said it is holding the dividend on its STRC preferred stock at 12%. The decision keeps the payout unchanged for investors in the company’s yield-focused security.
Read