OpenRouter Fusion

15 atoms · 跨 2 天 · 首见 2026-06-13 · 最近 2026-06-22

三色: 🟦 fact 6 · 🟥 take 9 · stance ▲6/▼0/◆0

来源: X 15

时态: fresh:15

标签: 好数字:7 · 好观点:5 · 好信源:5 · 好思考:1

叙事 (narrative) (1)

• 2026-06-22 · 第二个自适应编排API实例
  This is interesting, because it's the second instance, after OpenRouter Fusion, of an adaptive orchestration API. And I believe not the last, as this category will only grow from here.
  tags: 好观点 · → daily

预测 (forecast) (1)

• 2026-06-13 · 未来工作计划
  it calls for future work to benchmark both on long-horizon tasks
  tags: 好思考 · → daily

事实 (fact) (13)

• 2026-06-13 · 面板模型 vs 单模型表现
  Panels of models consistently outperform individual models
  tags: 好数字·好观点 · → daily
• 2026-06-13 · 预算模型组合能力
  Panels of budget models can surpass frontier models at a much lower cost
  tags: 好数字·好观点 · → daily
• 2026-06-13 · 预算面板性能对比
  the budget panel was comparable with Claude Fable 5 in performance
  tags: 好数字·好观点 · → daily
• 2026-06-13 · 融合性能提升来源比例
  roughly three quarters of the lift that Fusion provides comes from synthesis, and one quarter from diversity
  tags: 好数字 · → daily
  value: qty=~3/4 from synthesis, 1/4 from diversity
• 2026-06-13 · 合成性能提升幅度
  Synthesis lift: +6.7%
  tags: 好数字 · → daily
  value: qty=+6.7%
• 2026-06-13 · 多样性性能提升幅度
  Diversity lift: +2.1%
  tags: 好数字 · → daily
  value: qty=+2.1%
• 2026-06-13 · 预算面板与Claude Fable 5性能差距
  It landed within 1% of Fable 5 while costing roughly half the price
  tags: 好数字 · → daily
  value: qty=1% · direction=slightly below
• 2026-06-13 · DRACO基准测试范围
  We ran it on the DRACO deep research benchmark by Perplexity: 100 deep research tasks across 10 domains, from law and medicine to finance and product comparison
  tags: 好信源 · → daily
• 2026-06-13 · 基准测试数据处理
  We excluded those domains across every model with a one-line config change to the OpenRouter web search tool config, then re-ran everything. All published numbers come from the clean setup.
  tags: 好信源 · → daily
• 2026-06-13 · 测试时间
  We ran these benchmarks earlier in the week, before Fable was taken down
  tags: 好信源 · → daily
• 2026-06-13 · 成本对比方式
  Cost comparisons were performed including cache hits
  tags: 好信源 · → daily
• 2026-06-13 · 基准测试局限
  we have only evaluated one deep research benchmark so far, which did not include long-horizon tasks
  tags: 好信源 · → daily
• 2026-06-13 · 性能可达前沿水平
  Beyond-frontier performance can be achieved with frontier panels
  tags: 好观点 · → daily

时间轴 (近 20)

• 2026-06-22 · narrative · 第二个自适应编排API实例 · → daily
• 2026-06-13 · fact · 面板模型 vs 单模型表现 · → daily
• 2026-06-13 · fact · 性能可达前沿水平 · → daily
• 2026-06-13 · fact · 预算模型组合能力 · → daily
• 2026-06-13 · fact · 融合性能提升来源比例 · → daily
• 2026-06-13 · fact · 合成性能提升幅度 · → daily
• 2026-06-13 · fact · 多样性性能提升幅度 · → daily
• 2026-06-13 · fact · 预算面板性能对比 · → daily
• 2026-06-13 · fact · 预算面板与Claude Fable 5性能差距 · → daily
• 2026-06-13 · fact · DRACO基准测试范围 · → daily
• 2026-06-13 · fact · 基准测试数据处理 · → daily
• 2026-06-13 · fact · 测试时间 · → daily
• 2026-06-13 · fact · 成本对比方式 · → daily
• 2026-06-13 · fact · 基准测试局限 · → daily
• 2026-06-13 · forecast · 未来工作计划 · → daily

← 实体目录 · 系统日志