AI news · Monday, 26 October 2026

AI news, 26 October 2026

OpenAI slashes LLM prices up to 80%, launches Decisions API, while SpaceX secures $40 billion debt to buy AI GPUs.

23 stories, each labelled: available now, announced, unconfirmed or research. Get them ranked and explained for your product →

Paper of the week· 19 Oct – 25 Oct

BizBench: Unveiling Hidden Challenges in Deploying Long-Horizon Agents for Real-World Business

New benchmark shows top AI models fail over a third of the time on real business tasks due to process errors.

  1. Simulate Business Environment
  2. Define Complex Tasks
  3. Execute Agent Actions
  4. Validate Execution Path
  5. Calculate Strict Success
Read the 5-minute explainer

Top stories

Available now

OpenAI launches Decisions API, entering Jev decision‑model market

The Decisions API became generally available on October 22. It lets AI agents pick from predefined options, positioning OpenAI against TypeSafe’s Jev decision model.

Economic Times

SpaceX secures $40 billion debt financing to buy AI GPUs

Elon Musk posted on X that SpaceX has closed a $40 billion debt financing deal to purchase GPUs for AI workloads.

CNBC
Available now

OpenAI offers up to 90% discounts to poach Anthropic's biggest customers

Companies including Rippling, Replit and DataCamp say they cut AI costs by up to 90% after switching to OpenAI. Rippling raised OpenAI's share of its coding workload from 16% to 52% while Anthropic fell from 57% to 18%; Replit received more than 30% off a contract worth over $100 million a year, and DataCamp reported an 80‑90% cost reduction.

The Decoder
Available now

Hugging Face Releases Carbon‑X 14.8B Open DNA Language Model

Carbon‑X is a 14.8B DNA language model built on Qwen3‑14B, reads raw DNA in chat, answers in English, and averages 66.7% across five DNA benchmarks under an Apache 2.0 license.

MarkTechPost
Available now

Anthropic releases migration kit, shows single agent twice as fast as sub‑agents

Anthropic published a GitHub migration kit to help teams replace multi‑agent setups with a single agent that loads context on demand. Their tests show the single‑agent architecture runs about twice as fast as the multi‑agent approach.

The Decoder

Agents & dev tools