以下部分引用 AI 总结，现阶段 AI 仍然有幻觉。请以内容的原文为准。
数据新鲜度群聊 07-14 ✓卖方 07-14 ✓Wrap 断档15天 ✕News 07-15 ✓X 断档12天 ✕页面生成 07-15
GPT-Realtime-2
15 atoms · 跨 2 天 · 首见 2026-05-07 · 最近 2026-05-09
三色: 🟦 fact 15 · 🟥 take 0 · stance ▲7/▼0/◆5
来源: X 15
时态: fresh:15
标签: 好信源:15 · 好数字:11 · 好观点:1
事实 (fact) (15)
- 2026-05-07 ⭐ · Big Bench Audio benchmark score
achieving 96.6% in our Speech Reasoning benchmark, Big Bench Audio
tags: 好数字·好信源 · → daily
value: qty=96.6%
- 2026-05-07 ⭐ · Conversational Dynamics benchmark score #1
#1 in our Conversational Dynamics benchmark
tags: 好数字·好信源 · → daily
value: qty=96.1%
- 2026-05-07 ⭐ · context window size
model context window has increased from 32K to 128K
tags: 好数字·好信源 · → daily
value: qty=128K
- 2026-05-07 ⭐ · max output tokens
higher max output tokens (32k)
tags: 好数字·好信源 · → daily
value: qty=32k
- 2026-05-07 ⭐ · Conversational Dynamics benchmark minimal variant score
GPT-Realtime-2 is the leading model on Conversational Dynamics (Full Duplex Bench subset) benchmark with the minimal variant achieving a score of 96.1%
tags: 好数字·好信源 · → daily
value: qty=96.1%
- 2026-05-07 ⭐ · Conversational Dynamics benchmark high variant score
the High variant achieving a score of 95.3%
tags: 好数字·好信源 · → daily
value: qty=95.3%
- 2026-05-07 ⭐ · Time to First Audio on Big Bench Audio (high reasoning)
GPT-Realtime 2’s average Time to First Audio on the Big Bench Audio benchmark is 2.33 seconds on high reasoning
tags: 好数字·好信源 · → daily
value: qty=2.33 seconds
- 2026-05-07 ⭐ · Time to First Audio on Big Bench Audio (minimal reasoning)
1.12 seconds on minimal reasoning
tags: 好数字·好信源 · → daily
value: qty=1.12 seconds
- 2026-05-07 ⭐ · audio input price per hour
GPT-Realtime-2 audio input price remains steady at $1.15 per hour of audio input
tags: 好数字·好信源 · → daily
value: qty=$1.15
- 2026-05-07 ⭐ · audio output price per hour
$4.61 per hour of audio output
tags: 好数字·好信源 · → daily
value: qty=$4.61
- 2026-05-07 ⭐ · Big Bench Audio score increase from previous best
Model’s measured intelligence score on Big Bench Audio Speech to Speech reasoning benchmark of 96.6%, an increase of ~13% from previous highest result
tags: 好数字·好信源 · → daily
value: direction=increase · qty=~13%
- 2026-05-09 · 拥有工具使用能力
GPT-Realtime-2 has tool use.
tags: 好观点·好信源 · → daily
- 2026-05-07 · audio pricing unchanged
Audio pricing of model remains unchanged
tags: 好信源 · → daily
- 2026-05-09 · 已在主版本上可用
realtime 2 is on main and is fun to talk to on Discord or Google meet
tags: 好信源 · → daily
- 2026-05-09 · is for
GPT-Realtime-2 for instantly translating audio in realtime
tags: 好信源 · → daily
时间轴 (近 20)
- 2026-05-09 ·
fact · 拥有工具使用能力 · → daily
- 2026-05-09 ·
fact · 已在主版本上可用 · → daily
- 2026-05-09 ·
fact · is for · → daily
- 2026-05-07 ·
fact · Big Bench Audio benchmark score · → daily
- 2026-05-07 ·
fact · Conversational Dynamics benchmark score #1 · → daily
- 2026-05-07 ·
fact · context window size · → daily
- 2026-05-07 ·
fact · audio pricing unchanged · → daily
- 2026-05-07 ·
fact · max output tokens · → daily
- 2026-05-07 ·
fact · Conversational Dynamics benchmark minimal variant score · → daily
- 2026-05-07 ·
fact · Conversational Dynamics benchmark high variant score · → daily
- 2026-05-07 ·
fact · Time to First Audio on Big Bench Audio (high reasoning) · → daily
- 2026-05-07 ·
fact · Time to First Audio on Big Bench Audio (minimal reasoning) · → daily
- 2026-05-07 ·
fact · audio input price per hour · → daily
- 2026-05-07 ·
fact · audio output price per hour · → daily
- 2026-05-07 ·
fact · Big Bench Audio score increase from previous best · → daily
← 实体目录 · 系统日志