Tech Meridian ← LIVE FEED
PROMY MERIDIAN RU

NEWS · MODELS · #2448

OpenAI makes Decisions API generally available, repurposing Luna for fast decision tasks

OpenAI has made its Decisions API generally available to all customers, reconfiguring ChatGPT Luna for lower-latency, lower-cost decision (System One) use cases and adding voice and visual modalities. Initial customer trials (Box, Datadog, Ramp, Speak) show large latency reductions versus LLM calls, while benchmarks place OpenAI’s decision model behind some rivals on certain indices.

KEY POINTS

  1. OpenAI has made its Decisions API generally available to all customers, reconfiguring ChatGPT Luna for lower-latency, lower-cost decision (System One) use cases and adding voice and visual modalities.
  2. Initial customer trials (Box, Datadog, Ramp, Speak) show large latency reductions versus LLM calls, while benchmarks place OpenAI’s decision model behind some rivals on certain indices.
  3. This broad release brings a major provider into the emerging market for cheap, low-latency decision models, intensifying competition with specialized startups like TypeSafe and shaping fast, high-throughput AI use cases.
MERIDIAN INTELLIGENCE

DECISION BRIEF

70/100
CONFIDENCE
01

WHAT CHANGED

OpenAI made its Decisions API generally available and repurposed ChatGPT Luna to serve lower-latency, lower-cost decision (System One) workloads. The public release adds voice and visual modalities and OpenAI says it targeted inference resources geographically to lower latency.

02

WHY NOW

A major provider has entered the emerging market for cheap, low‑latency decision models, increasing competitive pressure on specialist vendors and enabling faster, higher‑throughput application flows that previously used LLM calls.

03

WHO IS AFFECTED

Developers and customers replacing LLM calls with faster decision models (explicitly: Box, Datadog, Ramp, Speak per trials), teams building voice/visual assistants, and operators focused on low‑latency, high‑throughput inference.

04

CONFIRMED

• OpenAI has made the Decisions API generally available to all customers. • OpenAI reconfigured ChatGPT Luna to reduce inference latency and cost for decision tasks rather than releasing a newly trained decision model. • The public Decisions API includes new voice and visual modalities and OpenAI targeted inference resources geographically to lower latency. • Initial trials with Box, Datadog, Ramp, and Speak reported large latency reductions versus LLM calls (TechCrunch cites overall drops from ~1–3 seconds to ~100–180 ms and Speak reporting 1.4s → 0.13s). • Public benchmarks named (AIM System One Index and Artificial Analysis Decision Index) currently favor other models such as Liquid’s D1 Ultra and Base Lab’s Elephant‑1. • OpenAI’s marketing describes the Decisions API as “Fast, focused, low‑cost decisions for any application.”

05

UNCERTAIN

Missing or conflicting evidence: • Exact technical changes applied to Luna (architecture, quantization, or training) beyond 'reconfiguration' are not described. • Benchmark methodologies and head‑to‑head scoring are not provided; public indices favor other models while customer trials report substantial latency gains, creating an apparent conflict in comparative performance. • Full pricing, throughput (requests/sec) and regional latency metrics are not published in the supplied excerpts. • Long‑term accuracy/calibration tradeoffs versus specialized decision models are not quantified. Label: these gaps are explicit in the sources or arise from lack of technical/benchmark detail in the excerpts.

06

WHAT TO WATCH

Observable signals to watch next (concrete): • Updates to AIM System One Index and Artificial Analysis Decision Index and any published head‑to‑head benchmark reports comparing OpenAI Decisions vs. Jev, D1 Ultra, Elephant‑1. • Official OpenAI release notes or technical posts describing whether Luna receives decision‑specific training or architecture/quantization changes. • Independently published latency/throughput measurements and measurement methodology from customers or third parties. • Competitor announcements and model releases (TypeSafe / Jev updates, Liquid D1 Ultra iterations, Base Lab) and resulting shifts in benchmark rankings. • OpenAI pricing, SLA or region availability changes for the Decisions API.

GROUNDED IN OpenAI ↗TechCrunch AI ↗

WHY IT MATTERS

This broad release brings a major provider into the emerging market for cheap, low-latency decision models, intensifying competition with specialized startups like TypeSafe and shaping fast, high-throughput AI use cases.

EVIDENCE MAP

4

Editorial claims linked to specific sources, with support, contradiction and context shown separately.

SOURCES & TIMELINE

2
01
OPENAI FIRST-PARTY
Build more responsive apps with the Decisions API

Fast, focused, low-cost decisions for any application.

↗
02
TECHCRUNCH AI INDEPENDENT COVERAGE
OpenAI makes its Jev competitor widely available as decision model competition heats up

OpenAI made its Decisions API model widely available to users for the first time, entering the market for fast, cheap intelligence that recently made TypeSafe, a startup founded by a former researcher at the frontier lab, a $7.6 billion company . TypeSafe’s Jev model created a category for deep learning models that act as super classifiers: Given a set of options, they can quickly make choices and offer a confidence…

↗