B.AI cuts Claude costs, adds GLM-5.2

// 1h agoINFRASTRUCTURE

B.AI cuts Claude costs, adds GLM-5.2

AI agent financial infrastructure platform B.AI has launched a Custom Provider feature and added Zhipu's GLM-5.2 model to its API gateway. The platform also introduced a free trial for MiniMax M3 and slashed Claude API pricing by up to 80% to lower costs for developers.

// ANALYSIS

B.AI is positioning itself as a highly flexible and cost-effective gateway for developers building autonomous AI agents.

–The Custom Provider feature allows developers to swap out default backends for any OpenAI-compatible API endpoint in coding tools.
–An 80% discount on the Claude series dramatically reduces the operational overhead of running token-heavy agentic workflows.
–The addition of GLM-5.2 and MiniMax M3 brings frontier-level 1M-token context window options to the platform's API gateway.
–Frictionless Web3-integrated billing allows autonomous agents to transact and pay for their own API usage without manual intervention.

// TAGS

b-aiglm-5-2minimax-m3llmagentapipricinginference

DISCOVERED

1h ago

2026-06-24

PUBLISHED

2h ago

2026-06-24

RELEVANCE

8/ 10

AUTHOR

KC_GMI

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

NEWS2h ago

Companies embed Cursor agents into tools

Companies are embedding Cursor's AI coding agents directly into their applications using the Cursor SDK. This integration provides users with a native agent experience directly within host platforms, bypassing the need to learn complex UIs.

BENCHMARK2h ago

VulcanBench completes first benchmark run

Morgan Linton completed the initial benchmark run of VulcanBench, an open-source evaluation tool for coding agents. While the test is complete, Linton notes that the 52 real-world coding tasks require further refinement to cover a wider difficulty spectrum.

UPDATE2h ago

LangChain, OpenRouter integration matures

Developers from both the LangChain and OpenRouter teams are celebrating the maturity of their dedicated integration, highlighting it as a highly reliable stack for AI applications. The first-party packages now support native tool calling, model reasoning tokens, and structured outputs without the need for manual workarounds.