TechDaily 科技早报

2026-09-14 · AI / 互联网 / 科技公司每日动态

今日 31 条新闻来自 6 个来源
实时资讯 →

模型能力排行 更新于 2026-09-14

综合能力 · AA 智能指数 — Artificial Analysis 智能指数(经 DataLearner 聚合) · 数据更新 2026-09-13 · 指标:智能指数 0-100

#模型机构分数
1Claude Fable 5.1Anthropic53
2Claude Fable 5.1Anthropic53
3GPT-6 Astra (max)OpenAI53
4GPT-6 Astra (xhigh)OpenAI53
5Claude Fable 5.1Anthropic51
6GPT-6 Astra (high)OpenAI51
7Claude Opus 5 (max)Anthropic51
8Claude Fable 5Anthropic50
9GPT-6 Astra (medium)OpenAI50
10Claude Opus 5 (xhigh)Anthropic50

对话体验 · LMArena 盲测 — LMArena 匿名人类盲测(经 DataLearner 聚合) · 数据更新 2026-09-11 · 指标:Elo 评分

#模型机构分数
1Claude Fable 5Anthropic1506
2Claude Opus 4.6 (high)Anthropic1505
3Opus 4.7 (high)Anthropic1502
4Claude Fable 5.1 (max)Anthropic1501
5F Muse Spark 1.2 (xhigh)Facebook AI研究实验室1499
6Claude Opus 4.6Anthropic1497
7Opus 4.7Anthropic1494
8F Muse Spark 1.1Facebook AI研究实验室1494
9Claude Opus 5 (high)Anthropic1493
10gemini-3.8-flash-highGoogle1493

代码能力 · SWE-bench Verified — SWE-bench 官方(swebench.com),真实 GitHub issue 工程任务 · 数据更新 官方持续更新 · 指标:500 题解决率 %

#模型机构分数
1Sonar Foundation Agent + Claude 4.5 Opus79.2%
2live-SWE-agent + Claude 4.5 Opus medium (20251101)79.2%
3TRAE + Doubao-Seed-Code78.8%
4live-SWE-agent + Gemini 3 Pro Preview (2025-11-18)77.4%
5EPAM AI/Run Developer Agent v20250719 + Claude 4 Sonnet76.8%
6Atlassian Rovo Dev (2025-09-02)76.8%
7Claude 4.5 Opus (high)76.8%
8ACoder76.4%
9Gemini 3 Flash (high)75.8%
10MiniMax M2.5 (high)75.8%

复杂推理 · Humanity's Last Exam — HLE 官方(lastexam.ai),CAIS + Scale AI,PhD 级难题 · 数据更新 官方持续更新 · 指标:准确率 %

#模型机构分数
1Gemini 3 Pro38.3%
2GPT-525.3%
3Grok 424.5%
4Gemini 2.5 Pro21.6%
5GPT-5-mini19.4%
6Claude 4.5 Sonnet13.7%
7Gemini 2.5 Flash12.1%
8DeepSeek-R1*8.5%
9o18.0%
10GPT-4o2.7%

今日要点

全部资讯 31

var rkTabs = document.querySelectorAll('.rk-tab'); rkTabs.forEach(function (t) { t.addEventListener('click', function () { rkTabs.forEach(function (x) { x.classList.remove('active'); }); document.querySelectorAll('.rk-panel').forEach(function (p) { p.classList.remove('active'); }); t.classList.add('active'); var el = document.getElementById(t.getAttribute('data-tab')); if (el) el.classList.add('active'); }); }); // ---------- 收藏(本地兜底 + Firebase 云同步) ---------- var FAV_KEY = 'techdaily_favs'; var favs = {}; function loadFavs() { try { favs = JSON.parse(localStorage.getItem(FAV_KEY)) || {}; } catch (e) { favs = {}; } } function saveFavs() { try { localStorage.setItem(FAV_KEY, JSON.stringify(favs)); } catch (e) {} } function favOf(link) { return favs[cleanLink(link)] || null; } function renderFavList() { var list = $('fav-list'); if (!list) return; var keys = Object.keys(favs).reverse(); list.innerHTML = keys.map(function (k) { var f = favs[k] || {}; return '
  • ' + '' + (f.title || '') + '' + '
    ' + (f.source || '') + '' + '' + (f.time || '') + '' + '
  • '; }).join(''); var sync = $('fav-sync'); if (sync) sync.textContent = '收藏保存在本机浏览器,不会上传到云端'; } function updateFavUI() { var n = Object.keys(favs).length; var c = $('fav-count'); if (c) c.textContent = n; var t = $('fav-total'); if (t) t.textContent = n; document.querySelectorAll('.fav-btn').forEach(function (b) { var on = !!favOf(b.getAttribute('data-link')); b.textContent = on ? '★' : '☆'; b.classList.toggle('active', on); }); var sec = $('favorites'); if (sec) sec.hidden = n === 0; renderFavList(); } function toggleFav(btn) { var link = cleanLink(btn.getAttribute('data-link')); if (!link) return; var li = btn.closest('.news-item'); var title = '', source = ''; if (li) { var tt = li.querySelector('.news-title'); if (tt) title = tt.textContent; var ss = li.querySelector('.src'); if (ss) source = ss.textContent; } if (favs[link]) { delete favs[link]; } else { favs[link] = { title: title, source: source, time: new Date().toLocaleString() }; } saveFavs(); updateFavUI(); } function initFavorites() { loadFavs(); updateFavUI(); } document.addEventListener('click', function (e) { var btn = e.target.closest('.fav-btn'); if (btn) { e.preventDefault(); e.stopPropagation(); toggleFav(btn); } }); var favNav = $('fav-nav-btn'); if (favNav) favNav.addEventListener('click', function () { var sec = $('favorites'); if (sec) { sec.hidden = !sec.hidden; if (!sec.hidden) sec.scrollIntoView({ behavior: 'smooth' }); } }); initFavorites(); // ---------- 回到顶部 + 全屏详情返回 ---------- var backTop = $('back-top'); if (backTop) { window.addEventListener('scroll', function () { backTop.classList.toggle('show', window.scrollY > 300); }, { passive: true }); backTop.addEventListener('click', function () { window.scrollTo({ top: 0, behavior: 'smooth' }); }); } var mBack = $('m-back'); if (mBack) mBack.addEventListener('click', closeModal); })();